JP2019536602A - Characterization and identification of anatomy - Google Patents

Abstract

The described embodiments relate to techniques for identifying and characterizing anatomy using machine learning techniques. These techniques are useful in certain types of tissues in anatomy, for example, among other structures, vascular (eg, vascular) tissues or lesions in animals (eg, human or non-human animals). And / or can be used to allow the machine to identify cells (eg, platelets, smooth muscle cells, or endothelial cells). Machine learning techniques may use the raw impedance spectroscopy measurement data in addition to the values resulting from this raw data. In addition, machine learning techniques can be used to select a frequency at which to measure the impedance, select a characteristic extracted from the measured impedance at the selected frequency, to arrive at a small set of frequencies that allows for reliable discrimination, Can be used. [Selection diagram] None

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This application is a pending US patent application Ser. No. 62/424, filed Nov. 21, 2016, entitled "CHARACTERIZING AND IDENTIFYING BIOLOGICAL Structure" filed on Nov. 21, 2016 under 35 U.S.C. 35 (119). , 693, the content of which is incorporated herein by reference in its entirety.

  Occlusion of blood vessels (including veins or arteries) can occur in various parts of an animal (eg, a human or non-human animal) and can have significant effects. For example, in an ischemic stroke, a blood clot completely or partially inhibits cerebral artery blood flow. If the clot is not processed quickly, insufficient blood flow may cause irreparable damage to the brain.

  The obstruction may be due to a blood clot that may result from the clotting of red blood cells and / or white blood cells and / or platelets in the blood vessel. Clotting can be caused by a variety of factors, including damage, abnormal blood flow at the site of the obstruction, diseases / symptoms that make the animal more likely to clot, and / or other factors.

  A common clot treatment is chemical lysis of the clot, which is feasible within the first 4.5 hours after occlusion of a blood vessel. Another common option is mechanical thrombus removal, where a suction catheter or stent-type thrombectomy is used to remove blood clots from blood vessels.

  Stent clot collectors include a stent attached to the end of a wire. The stent deploys in the vasculature and in the clot, expands in the clot, and is withdrawn, usually after a waiting period of 0.5-10 minutes, to pull the clot out of the blood vessel. Retrieving a clot with a stent-based thrombus collector in a sub-optimal condition may result in several consecutive treatments (on average three times) because some of the clot may remain or be lost from the collector. May be necessary to treat occlusions and to restore vascular circulation. Each repetition increases the damage to the vessel wall, increasing both the duration of the intervention and the period in which blood flow is impeded by the obstruction, and can result in irreversible animal damage. The physio-mechanical process of clot retrieval is currently poorly understood, but the two most common explanations for suboptimal clot retrieval are: (1) stent-type thrombus The collector does not deploy in the clot, only the friction induced by the stented collector pressing the clot against the wall contributes to the clot recovery, and (2) the stent It has been described that although deployed into the clot, an insufficient amount of time was provided for the stent to coalesce with the clot.

  If a suction catheter is used to remove a clot, the clinician inserts the catheter into the vasculature and manipulates the catheter to aspirate the clot into the catheter. Depending on the diameter of the catheter, the catheter may be placed in direct contact with the clot, or may be placed in the area of the proximal blood vessel. Depending on the composition and viscosity of the clot, the suction method may be different. Some difficulties can arise when using a suction catheter. For example, if a blood clot is aspirated into a catheter, the aspiration may be obstructing flow through the catheter. In such a situation, the clinician may not know if the clot is obstructing the tip of the catheter or if it is inside the catheter and obstructing the tube if the catheter is not withdrawn. If a blood clot is obstructing the tip of the catheter, the blood clot can travel through the bloodstream and squeeze the blood vessels in another part of the animal by the unknowing release of the blood clot during removal of the catheter. There is a risk of obstructive embolism.

  The described embodiments relate to techniques for identifying and characterizing anatomy using machine learning techniques. These techniques are useful in certain types of tissues and / or tissues in anatomy, such as vascular (eg, vascular) tissues or lesions in animals (eg, human or non-human animals), among other structures. Or it can be used to allow the machine to identify cells (eg, platelets, smooth muscle cells, or endothelial cells). Machine learning techniques may use the raw impedance spectroscopy measurement data in addition to the values resulting from this raw data. In addition, machine learning techniques select a frequency at which to measure the impedance, select a characteristic extracted from the measured impedance at the selected frequency, to arrive at a small set of frequencies that allows for a reliable discrimination. Can be used.

  In one embodiment, a method is provided for training a system to identify at least one characteristic of a anatomy. In some embodiments, the method includes receiving training data including a plurality of sets of impedance measurements for the anatomy, and including a first subset of impedance measurements from each set of the plurality of sets of impedance measurements. Identifying a first subset of the training data; and identifying a first plurality of characteristics from the first subset of the identified training data, wherein the first plurality of characteristics comprises the identified training data. Training a model using at least one machine learning technique with a first plurality of identified properties to include at least one derived property arising from a first subset of Creating a model.

  The described embodiments are useful when inserted into an animal (eg, a human or non-human animal, including a human or non-human mammal), such as a vasculature that completely or partially inhibits a vessel. Medical devices, including invasive probes, that may be useful for the diagnosis and / or treatment of tumors or deposits in vessels). An invasive probe may include one or more sensors for detecting features of the lesion, including detection by detecting one or more features of the tissue and / or biological material of the lesion. The medical device may be configured to analyze the characteristics of the lesion and provide a clinician with a treatment recommendation based on the analysis. Such treatment recommendations may include how to treat the lesion, for example, how to use which treatment to treat the lesion, and / or how to use a treatment device. In some cases, the subject matter of the present invention includes correlated products, alternative solutions to specific problems, and / or multiple different uses of one or more systems and / or articles.

  In one embodiment, a medical device for diagnosing and / or treating a lesion in an animal is provided. The medical device includes an invasive probe for insertion into an animal and removal from the animal after diagnosis and / or treatment, the invasive probe comprising at least one sensor, at least one processor, and at least one processor Includes at least one storage medium having executable encoded instructions for causing at least one processor to perform a method. The method includes identifying a composition of the lesion using at least one sensor, wherein identifying the composition of the lesion comprises determining one or more biological materials present in the lesion and determining the composition of the lesion. Identifying at least one characteristic of the lesion based at least in part on the lesion.

  In certain embodiments, the medical device is an invasive probe configured to be inserted into an animal tube during diagnosis and / or treatment of a vascular lesion and removed from the tube after the diagnosis and / or treatment. Wherein the invasive probe is configured to perform one or more measurements of a vascular lesion, wherein the invasive probe comprises at least one impedance sensor and at least one impedance sensor. At least one circuit for making the plurality of measurements of impedance, wherein each measurement of the plurality of measurements of impedance corresponds to one of the plurality of frequencies, and wherein the electrical signal at the corresponding frequency is When applied to a lesion, it is a measure of the impedance of the lesion.

  Certain aspects relate to methods of the present invention for operating a medical device for diagnosing and / or treating a lesion in an animal. The medical device includes an invasive probe to be inserted into the animal and removed from the animal after diagnosis and / or treatment of the lesion. The method includes using an invasive probe of a medical device while an invasive probe is placed in an animal, and measuring impedance spectra of a plurality of biomaterials of the lesion as measured by the invasive probe at multiple locations of the lesion. Creating a digital signal that represents operating the invasive probe to apply an electrical signal at a plurality of frequencies and measuring impedance of a plurality of biological materials of the lesion. Actuating a plurality of sensors of the invasive probe to The method further includes identifying the lesion based at least in part on the analysis of the digital signal, and using at least one processor of the medical device to at least partially analyze the digital signal and / or the identity of the lesion. Determining one or more treatment recommendations for a method of treating a lesion, and outputting one or more treatment recommendations for presentation to a user via a user interface.

  In a further embodiment, a method of operating a medical device for diagnosing and / or treating a lesion in an animal, wherein the medical device is inserted into the animal and removed from the animal after diagnosis and / or treatment of the lesion. A method comprising an invasive probe. The method uses the invasive probe of the medical device while the invasive probe is placed in the animal to generate data representative of one or more electrical properties of the biological material present in the animal's lesion. Generating the data comprises activating at least one sensor of the invasive probe to measure one or more electrical properties of the biological material present in the lesion; and via a user interface And outputting information representing one or more electrical properties for presentation to a user.

  In some embodiments, an apparatus is described. In certain embodiments, the apparatus includes at least one processor and at least one storage having executable encoded instructions that, when executed by the at least one processor, cause the at least one processor to perform a method. Medium, the method comprising receiving, over time, reports from a plurality of medical devices on medical treatment performed on a plurality of lesions in a vessel of an animal, wherein each report of the plurality of reports comprises Including one or more characteristics of the lesion treated with the medical treatment, one or more parameters of the corresponding medical treatment performed to treat the lesion, and prognostic indicators for the corresponding medical treatment; One or more relationships between parameters of successful lesion treatment and / or unsuccessful lesion treatment in multiple reports on healthcare Learning over time, wherein learning the one or more relationships includes determining one or more conditions associated with each treatment option of the plurality of treatment options; If one or more conditions satisfy one or more conditions for the treatment option, the corresponding treatment option is associated with the lesion feature as recommended for the treatment of the lesion. Multiple medical devices to make recommendations to the clinician among multiple treatment options, based on an assessment of lesion characteristics with respect to one or more conditions associated with each of the multiple treatment options. Setting.

  At least one storage medium having executable encoded instructions that, when executed by at least one processor, causes the at least one processor to perform a method has been described according to certain embodiments. In some embodiments, the method comprises receiving, over time, a plurality of reports regarding medical care performed on the plurality of lesions in the animal tract from the plurality of medical devices, wherein each report of the plurality of reports is provided. Comprises one or more characteristics of the corresponding medically treated lesion, one or more parameters of the corresponding medical treatment performed to treat the lesion, and a prognostic indicator for the corresponding medical treatment; Learning one or more relationships between the characteristics of the disease and parameters of successful and / or unsuccessful treatment of the lesion over time based on medical reports. Learning the above relationship includes determining one or more conditions associated with each treatment option of the plurality of treatment options, wherein the one or more conditions are associated with a treatment option corresponding to a feature of the lesion. About One or more associated with each of the multiple treatment options, as the corresponding treatment option is recommended for the treatment of the lesion if one or more conditions are met; Setting a plurality of medical devices to make a recommendation to a clinician among a plurality of treatment options based on the evaluation of the characteristics of the lesion with respect to the condition.

  Particular embodiments are receiving, over time, medical reports on medical treatments performed on multiple lesions of an animal tract, wherein the reports of the plurality of reports are associated with corresponding medical treatments. Including one or more characteristics of the lesion treated in the above, one or more parameters of the corresponding medical treatment performed to treat the lesion, and a prognostic indicator for the corresponding medical treatment; Learning one or more relationships between parameters of successful and / or unsuccessful treatments over time based on applying a machine learning process to a plurality of reports on healthcare, Learning the one or more relationships includes determining one or more conditions associated with each treatment option of the plurality of treatment options, wherein the one or more conditions correspond to features of the lesion. If one or more of the treatment options are met, the corresponding treatment option is related to the characteristics of the lesion, as recommended for treatment of the lesion; and is associated with each of the multiple treatment options Performing at least one action, such as setting up multiple medical devices, to make recommendations to a clinician among a plurality of treatment options based on an assessment of the characteristics of the lesion with respect to one or more conditions A method is described that includes operating a processor.

  In a further embodiment, a method of diagnosing and / or treating a lesion in an animal, comprising inserting an invasive probe of a medical device into the animal, wherein the invasive probe comprises a plurality of biomaterials each of the lesion. Including at least one sensor for measuring one or more characteristics of the lesion; identifying the lesion based at least in part on one or more characteristics of each of the plurality of biological materials of the lesion; Operating the medical device to make one or more recommendations for treatment of the lesion based at least in part on one or more characteristics and / or identity of each of the plurality of biological materials; Methods are provided that include treating a lesion according to one or more recommendations of a medical device and removing an invasive probe from a vessel of an animal.

  In some embodiments, a medical device configured to diagnose and / or treat vascular lesions in an animal is described. In certain embodiments, the medical device includes inserting an invasive probe of the medical device into a tube of an animal, wherein the invasive probe measures one or more characteristics of the tissue and / or biological material of the lesion. And at least one sensor based on at least one measurement of the at least one sensor of the invasive probe to generate one or more recommendations for treatment of the lesion. And configured to deliver treatment to the lesion in accordance with one or more recommendations for treating the lesion. In certain embodiments, the medical device is also configured to remove a lesion from a tube of an animal.

  In one embodiment, a method is provided for training a system to identify at least one characteristic of a anatomy. The method includes receiving training data that includes a plurality of sets of impedance measurements for a anatomy, and, from each set of the plurality of sets of impedance measurements, a first subset of training data that includes a first subset of the impedance measurements. And identifying a first plurality of characteristics from the first subset of the identified training data, wherein the first plurality of characteristics results from the first subset of the identified training data. Including at least one derived property, training the model using at least one machine learning technique with a first plurality of identified properties to create a first trained model; ,including.

  In another embodiment, a method is provided for training a system to identify at least one characteristic of a anatomy. The method comprises operating at least one processor to select a subset of impedance measurements from each of the plurality of sets of impedance measurements for the anatomy and to perform an action such as generating the plurality of subsets of impedance measurements. Including. Each set of impedance measurements includes one impedance measurement of the anatomy in response to application of a different frequency signal. The method further includes operating the at least one processor to perform an action such as creating a plurality of sets of characteristics. Each set of properties characterizes a subset of the impedance measurements of the plurality of subsets. Each set of properties includes at least one property present in the subset of impedance measurements and at least one derived property resulting from the subset of impedance measurements. Further, the method operates at least one processor to perform an action, such as training a model to recognize at least one feature of the anatomy of interest, based on the input impedance measurement for the anatomy of interest. including. The training includes applying at least one machine learning technique to sets of properties that characterize subsets of impedance measurements to create a trained model.

  In a further embodiment, a method is provided for training a system to identify at least one characteristic of a anatomy. The method may include at least performing a plurality of sets of impedance measurements, and in each set of impedance measurements, performing an action such as training a first model using the anatomical index corresponding to the set of impedance measurements. Including activating one processor. The multiple sets of impedance measurements include impedance measurements for multiple types of anatomy. The training trains the first model based at least in part on the impedance measurements to distinguish impedance measurements for the first type of anatomy from impedance measurements for one or more other types of anatomy. Including doing. The training includes applying at least one machine learning technique. The method further includes applying at least one machine learning technique to the impedance measurement for the first type of anatomy at least in part to identify at least one characteristic of the first type of biological material. Activating at least one processor to perform an action, such as training two models.

  In another embodiment, a method is provided for determining at least one characteristic of a anatomy. The method includes operating the at least one processor to perform an action, such as evaluating an impedance measurement for the anatomy using the at least one trained model to determine at least one characteristic. . At least one trained model is trained to distinguish between anatomical structures having different characteristics.

  In a further embodiment, a method is provided for determining a method of treating a pathology in an animal. The method includes operating the at least one processor to perform an action, such as evaluating an impedance measurement for the lesion using the at least one trained model to determine how to treat the lesion. . Evaluating the impedance measurement includes evaluating one or more characteristics of the impedance measurement using at least one trained model. The one or more properties include at least one derived property that results from the impedance property.

  In another embodiment, a method is provided for training a system to identify a method for treating a anatomy. The method includes receiving training data including a plurality of sets of impedance measurements for a anatomy, and, from each set of the plurality of sets of impedance measurements, a first subset of training data including a first subset of impedance measurements. And identifying a first plurality of characteristics from the first subset of the identified training data, wherein the first plurality of characteristics results from the first subset of the identified training data. Including at least one derived property, training the model using at least one machine learning technique with a first plurality of identified properties to create a first trained model; ,including.

  In a further embodiment, there is provided a method of training a system to identify a method of treating a anatomy. The method comprises operating at least one processor to select an impedance measurement subset from each of the plurality of sets of impedance measurements for the anatomical structure and to perform an action such as generating the plurality of impedance measurement subsets. including. Each set of impedance measurements includes one impedance measurement of the anatomy in response to application of a different frequency signal. Further, the method further includes operating the at least one processor to perform an action such as creating a plurality of sets of characteristics. Each set of properties characterizes a subset of the impedance measurements among the plurality of subsets. Each set of properties includes at least one property present in the subset of impedance measurements, and at least one derived property resulting from the subset of impedance measurements. Further, the method trains the model to perform at least one action, such as determining from input impedance measurements for the anatomy of interest, a recommended treatment for the anatomy of interest among a plurality of treatment options. Operating one processor. The training includes applying at least one machine learning technique to sets of properties that characterize subsets of impedance measurements to create a trained model.

  In another embodiment, a method is provided for training a system to identify a method for treating a anatomy. The method may include at least performing a plurality of sets of impedance measurements, and in each set of impedance measurements, performing an action such as training a first model using the anatomical index corresponding to the set of impedance measurements. Including activating one processor. The multiple sets of impedance measurements include impedance measurements for multiple types of anatomy. The training trains the first model based at least in part on the impedance measurements to distinguish impedance measurements for the first type of anatomy from impedance measurements for one or more other types of anatomy. Including. The training includes applying at least one machine learning technique. The method further comprises training at least one machine learning technique at least in part on the impedance measurement for the first type of anatomy to train the second model from among a plurality of treatment options. Activating the at least one processor to perform an action, such as determining a recommended treatment for the first type of anatomy of interest.

  In a further embodiment, there is provided at least one storage medium having executable encoded instructions that, when executed by at least one processor, causes at least one processor to perform any of the methods described above.

  In another embodiment, at least one processor and at least one having executable encoded instructions that, when executed by the at least one processor, cause the at least one processor to perform any of the above methods. And a storage medium.

  Other advantages and novel features of the present invention will become apparent from the following detailed description of various non-limiting embodiments of the invention, when considered in conjunction with the accompanying drawings. In cases where the present specification and a document incorporated by reference include conflicting and / or inconsistent disclosure, the present specification controls. Thus, the above is a non-limiting summary of the invention as defined by the appended claims.

  The accompanying drawings are not intended to be drawn to scale. In the drawings, each identical or nearly identical component that is illustrated in various figures is represented by a like numeral. For clarity, not all components may be named in all drawings.

FIG. 9 is a flowchart of a method by which a clinician can operate a medical device for diagnosing and / or treating a lesion, according to embodiments described herein. 9 is an illustration of an example of a medical device according to some embodiments. 9 is an illustration of an example of an invasive probe according to some embodiments. FIG. 4 is a flowchart of a process that may be performed in some embodiments to determine the composition of a lesion. FIG. 4 is a flowchart of a process that may be performed in some embodiments to determine the composition of a lesion. 5 illustrates an absolute value of an exemplary frequency spectrum of lesion impedance. FIG. 5 illustrates an exemplary model of the impedance of a lesion that can be implemented in the method of FIG. 4, including a constant phase element. FIG. 5 illustrates an exemplary model of the impedance of a lesion that can be implemented in the method of FIG. 4, including a constant phase element. FIG. 5 illustrates an exemplary model of the impedance of a lesion that can be implemented in the method of FIG. 4, including a constant phase element. FIG. 5 illustrates an exemplary model of the impedance of a lesion that can be implemented in the method of FIG. 4, including a constant phase element. 5 illustrates an exemplary system for implementing the method of FIG. FIG. 5 is a flowchart of an exemplary method for operating a medical device according to some embodiments described herein for generating a treatment advisory. FIG. 11 is a flowchart of another exemplary method of some embodiments for operation of a medical device according to embodiments described herein for generating a treatment recommendation based in part on a lesion composition. 5 is a flowchart of an exemplary method of creating a treatment recommendation using conditions, which may be performed in some embodiments. FIG. 9 is a flowchart of an exemplary process for operating a server to analyze a report on a treatment to determine conditions for setting a medical device, which may be implemented in some embodiments. FIG. 9 is a flowchart of an exemplary process for operating a server to analyze a report on a treatment to determine conditions for setting a medical device, which may be implemented in some embodiments. FIG. 9 is a flowchart of an exemplary process for operating a server to learn the relationship between raw data and options for successful treatment, which may be implemented in some embodiments. 5 is an example of a process that may be performed in some embodiments for creating a treatment chronicle. 9 is an example of a process that may be performed in some embodiments to train a model to accurately characterize a lesion and / or to provide treatment recommendations. 5 is an example of a process that may be implemented in some embodiments for creating a filter using a machine learning model. 9 is an example of a process that may be implemented in some embodiments to assist a clinician in guiding an invasive probe. 5 shows, in diagrammatic form, an example of the effective capacitance of a cell structure determined by the method of FIG. 1 illustrates an example of a system created according to aspects of the present disclosure. 1 illustrates an example of a system created according to aspects of the present disclosure. 4 is a histogram showing the effective capacitance of multiple types of cells determined under controlled conditions. 4 is a histogram showing the effective capacitance of multiple cell types determined under uncontrolled conditions. 5 is a flowchart of an exemplary process for training a model using machine learning techniques to analyze biological material. 23 is an example of a confusion matrix resulting from applying training data to a trained model created by the process illustrated in FIG. 23 is an example of a confusion matrix resulting from applying test data to a trained model created by the process illustrated in FIG. 23 is an example of a confusion matrix resulting from applying test data to a trained model created by the process illustrated in FIG. 9 is a graph showing amplitude and phase spectra for experimental data. 9 is a graph showing amplitude and phase spectra for experimental data. 9 is a histogram showing various parameter distributions. 9 is a histogram showing various parameter distributions. 9 is a histogram showing various parameter distributions. 9 is a histogram showing various parameter distributions. 9 is a histogram showing various parameter distributions. 9 is a histogram showing various parameter distributions. 9 is a histogram showing various parameter distributions. 9 is a histogram showing various parameter distributions. 4 is a histogram showing the distribution of values representing the effective capacitance of different cell types. 4 is a histogram showing the distribution of values representing the effective capacitance of different cell types. 4 is a histogram showing the distribution of values representing the effective capacitance of different cell types. Exemplary embodiments of some embodiments for operating a medical device according to embodiments described herein to generate a treatment recommendation based in part on characteristics of cancerous and / or non-cancerous tissue. 6 is a flowchart of a method. Exemplary embodiments of some embodiments for operating a medical device according to embodiments described herein to generate a treatment recommendation based in part on characteristics of cancerous and / or non-cancerous tissue. 6 is a flowchart of a method. Exemplary embodiments of some embodiments for operating a medical device according to embodiments described herein to generate a treatment recommendation based in part on characteristics of cancerous and / or non-cancerous tissue. 6 is a flowchart of a method. FIG. 4 is a block diagram of a computing device on which some embodiments may operate.

  The application file of U.S. Provisional Patent Application No. 62 / 424,693, to which this application claims priority, includes at least one of the above drawings, depicted in color. Copies of the color drawings will be provided by the US Patent and Trademark Office after request and payment of the necessary fee.

DETAILED DESCRIPTION Certain embodiments described herein provide for the treatment of animal lesions when implanted or inserted into an animal (eg, a human or non-human animal, including a human or non-human mammal). The present invention relates to medical devices that include invasive probes that can be useful for diagnosis and / or treatment. The lesion can be an abnormality in the anatomy of the animal, such as a deviation from normal structure, and / or an abnormality in the function of some of the animal, such as an abnormality associated with an injury, medical condition, or disease. Lesions may appear in different parts of the animal, for example, may be contained in the animal's tract. Vascular lesions can act, for example, as an obstruction that completely or partially obstructs the vessel. The tract may be, for example, a blood vessel or other tract of an animal, but the lesion may be wholly or partially formed by a tumor in the tract, accumulation of material in the tract, and / or other causes of the lesion. Is also good. An invasive probe may include one or more sensors for detecting features of the lesion, which may include detecting the composition of the lesion.

  In some embodiments, detecting the composition of the lesion comprises removing one or more cells and / or one or more tissues present in the lesion and / or one or more plaque-like substances present in the lesion. And identifying one or more biological materials of the lesion, including. The biological material of the lesion to be identified may be all the biological material present in the lesion or only a part of the biological material present in the lesion. If only a portion of the biological material has been identified, then the identified material is simply a particular type of material, such as diseased tissue / cells (as compared to other materials, such as plaque-like materials). Or may be a particular type of tissue / cell (eg, red blood cells present in a lesion, not other types of cells). If the composition has been determined and if only one or several types of biological material have been identified, determining this composition may involve determining the amount of the identified substance in the lesion, for example, It may include determining the amount of the identified substance relative to all the substances in the lesion, including by calculating the ratio of one or more identified substances to all substances.

  In some such embodiments, an invasive probe may identify and / or classify a lesion. Identifying or classifying a lesion may, in some embodiments, include diagnosing the lesion. The medical device may be configured to analyze the lesion and provide a clinician with a treatment recommendation based on the analysis. Such treatment recommendations may include methods of treating the lesion, such as which treatment or combination of treatments to use to treat the lesion (eg, if a lesion is to be removed, a suction catheter or stent-type thrombus). And / or how to use the treatment device (how fast the stent-type thrombus collector is withdrawn). In some such embodiments, the invasive probe is at least associated with the composition of the lesion (eg, the identity of one or more biological materials of the lesion), and / or one or more other characteristics of the lesion as a whole. Based in part, such identification or classification may be made and / or the treatment recommendations may be made. For example, a system that includes an invasive probe and / or other related devices described below can identify the biological material of the lesion and identify the type of lesion based on the identified material, or otherwise. Lesions can be classified. Based on the identification of the lesion, the system may make recommendations for the treatment of a particular type of lesion. As another example, in other embodiments, the system may identify the biological material of the lesion and make recommendations regarding the treatment of the lesion based on the identified material (alone or in comparison to other substances in the lesion). I can do it. As another example, in other embodiments, the system may make recommendations regarding treatment of a lesion based on information characterizing the biological material of the lesion (eg, impedance spectrum).

  Additionally or alternatively, embodiments relate to techniques for identifying and characterizing anatomy using machine learning techniques. These techniques are particularly useful in anatomical structures, such as vascular (eg, vascular) tissues or lesions in animals (eg, human or non-human animals), among other structures. And / or may be used to allow the machine to identify cells (eg, platelets, smooth muscle cells, or endothelial cells). Machine learning techniques may use raw impedance spectroscopy measurement data in addition to values derived from this raw data. In addition, machine learning techniques can be used to select the frequency at which to measure impedance and arrive at a small set of frequencies that allows for reliable discrimination.

  In some embodiments, an invasive probe may include one or more sensors, which may include sensors for measuring the impedance of a lesion. The sensor may measure the impedance of the lesion when applying an electrical signal having a particular frequency to the lesion. The medical device may be configured to determine a composition of the lesion and / or one or more characteristics of the lesion based on the value of the impedance. For example, each sensor may, in some embodiments, detect an impedance spectrum of a biological material in contact with the sensor, such that different sensors of the invasive probe may simultaneously generate different impedance spectra for different biological materials of the lesion. Can be operated. Thus, the medical device may make a treatment recommendation based in part on the determined composition. As described above, determining the composition may include identifying an amount of one or more biological materials in the lesion, which may be less than all the materials in the lesion. For example, in some embodiments, the amount of a lesion made up of red blood cells is determined.

  The various examples described herein discuss medical devices and methods of treating vascular lesions associated with vascular lesions. However, it should be understood that embodiments are not so limited. The techniques described herein for detecting features of a lesion and generating a treatment recommendation may be performed at any suitable lesion in a duct on the anatomy of the animal or at other locations within the anatomy of the animal. It can be used with any suitable lesion, including possible lesions. Where the lesion is a vascular lesion, such vessels may include, for example, vascular and digestive tracts. One skilled in the art will appreciate that anatomical tubes are different from anatomical cavities. For example, a tube may be significantly smaller in one dimension (eg, width) than another dimension (eg, length).

  Thus, in some embodiments, an invasive probe may be a component of a medical device for diagnosing and / or treating vascular pathologies. For example, the medical device may be a thrombectomy device, and the invasive probe may be a component of the thrombectomy device. Thus, an invasive probe can be a component of a guidewire, suction catheter, microcatheter, stent-retriever, and / or another thrombectomy device. In some embodiments, the medical device may include two or more of a guidewire, a suction catheter, and a stent-type thrombus collector, and the invasive device may include one or more of these (all of them). May be included).

  We have found that conventional general techniques for lesion identification, including identifying the type of vascular lesion based on electrical measurements of the lesion, are not sufficient for effective use in medical institutions. Recognized and understood not to be accurate or reliable. Some such prior art techniques involve creating multiple impedance spectra for each of the various types of lesions and for the entire lesion, as well as creating an “average” impedance spectrum for each type of lesion. However, lesions vary widely from person to person or within the same person, making it possible to achieve an accurate or representative "average" or "standard" impedance spectrum for the lesion as a whole. Making it difficult. Other such conventional techniques impose a rigorous measurement process, thereby allowing precise placement of the sensor in contact with the lesion for each measurement during the determination of the "standard" spectrum and subsequent use in the patient. Attempting to improve reliability by requiring the same measurement location. Such precise placement is almost impossible to replicate and reproduce in practice and is still reliable to the point where these techniques can be used with sufficient accuracy to be useful in patient use Did not improve. For example, during use on a patient, measurements of the impedance spectrum of the patient's lesions need to be made, and the measurements need to be compared against multiple "standard" spectra for each type of lesion, and the Statistical analysis needs to be performed to identify the type of lesion. However, with typical prior art, even these combined analyzes result in a degree of certainty of at most just over 50%.

  Some embodiments described herein are directed to a method of identifying a type of lesion, the method using a database of "standard" impedance spectra for the lesion as a whole, or such a whole lesion. Does not include statistical analysis to compare the impedance spectra of Some of these embodiments are configured to characterize a lesion by identifying the composition of the lesion, such as the type and number of some or all of the biological material present in the lesion. This includes identifying one or more tissues and / or cells of the lesion and / or one or more plaque-like substances present in the lesion. Thus, in some such embodiments, the composition of the lesion may be analyzed to identify the features of the lesion with a high degree of certainty. Such lesion characteristics may include the type of lesion, and embodiments may include diagnosing the lesion. In some embodiments, the composition of the lesion is a condition that identifies that a particular type of lesion is associated with a particular composition (eg, a particular set of biological materials, or a particular relative amount of biological material). Can be compared to one or more conditions related to the type of lesion. If a particular composition is determined to be consistent with a lesion type by satisfying conditions associated with the lesion type, then a lesion having this composition may be identified as a lesion of that type. The identification of a lesion based on the identification of the biological material of the lesion may have high reliability (eg, greater than 90%).

  In addition, the inventors have found that conventional medical devices, including conventional thrombectomy devices, do not provide information about the characteristics of lesions of the vasculature, including blood vessels, and that conventional medical devices do Recognize and understand not to provide. We further recognize and understand that this lack of information is a cause of the difficulty in treating the lesion. For example, without using information about the composition of the lesion, it may be difficult for the clinician to choose among the available treatment options, as each treatment option may have a different composition of lesion. It may work best. In addition, without using information about the status of treatment of the lesion, the clinician may not be aware that the treatment was successful or not. Due to the lack of this information, multiple treatments may be needed to correctly treat the lesion. Each such treatment increases the risk of injury to the patient, and more importantly, for some lesions, increases the duration of the lesion. If blood vessels are partially or completely inhibited by the lesion, reduced blood flow can cause damage to animal tissues.

  Thus, in the embodiments described herein, the medical device can determine the characteristics of the lesion, monitor the performance of the treatment, and provide recommendations regarding how to treat the lesion before and / or during treatment. Can create. This additional information may be useful in treating lesions in an effort to ensure that the lesion is removed with only one treatment and that subsequent treatment is not necessary for the same lesion, or at least to increase the chance. Clinicians can be assisted in tentatively determining how to do so and in providing treatment. The medical device may provide information to the clinician in real time during the medical procedure, for example, by providing the clinician with real-time information regarding the interaction between the medical device and the lesion. Real-time, in some embodiments, includes providing a clinician with information of the corresponding data detected by the medical device within a period of time, where the time is less than 5 seconds, less than 10 seconds, 30 seconds or less. It may be less than a second, less than a minute, or less than 5 minutes, which may depend on the requirements of the analysis performed on the data to make recommendations.

  In some embodiments, the reliability and effectiveness of the techniques and devices can be improved through initialization and configuration, which can be useful when characterizing the material of the anatomy or the structure itself, and / or treating Includes improvements by setting up the system using a particular approach to the selection of data points to be collected, created, and / or used with respect to the anatomy used in making the recommendations. Determining such collected data points may include determining the frequency at which the impedance spectrum of the biological material is measured. Additionally or alternatively, after retrieving the impedance values for the frequency (e.g., a range of frequencies or any set of selected frequencies), any of the characteristics of these impedance values should be used for subsequent analysis. This includes determining which of the distinct data values in the set of impedance values to use, or what values are derived from analyzing the data values, for use in subsequent analyses. Including identifying what should be done. Some embodiments may include identifying such frequencies and / or characteristics using machine learning analysis.

  Properties in which embodiments may operate include descriptors or attributes of data or sets of data, including descriptors or attributes of sets of impedance measurements. A descriptor or attribute of a measurement or set of measurements may characterize a data point or set of data of the data set. A property may have a value, such as a numerical value. If the characteristics are used to characterize different measurements or datasets of measurements, the characteristics may be the same descriptor or attribute in different measurements / sets, and thus characterize the measurements / sets in the same or similar manner. , But may have different values corresponding to these different measurement / set data points.

  Properties of the type described herein may include properties that are present in the impedance measurement or set of impedance measurements (eg, electrical impedance spectroscopy measurements) and / or properties that result from the impedance measurement or set. Properties present in or derived from the impedance measurement may include numerical values explicitly set in the measurement or measurement data set. Examples of such properties are the magnitude or phase of the impedance measurement, or the numerical value of the minimum or maximum value of the set in the impedance measurement set (where the minimum / maximum values are absolute and / or relative values). Value). The minimum or maximum data value may require some analysis, such as comparing values, but the minimum / maximum value itself is the value found in the data set. Derived properties may describe the measurement or set, but may include values not found in the measurement / set. Alternatively, the value of the derived property may result from the measurement / set and may be obtained, for example, via performing one or more computer processes on the measurement / set. Examples of derived properties include, among others, the average value of the EIS measurement, the phase maximum frequency of the EIS measurement, the n quantile of the EIS measurement, the first derivative of the EIS measurement, and the second derivative of the EIS measurement. Functions.

  In some embodiments, the initialization and settings distinguish between a target anatomy, such as a particular type of tissue or a particular type of lesion, and impedance measurements and / or properties associated with other anatomy. Training the filter for In some scenarios, when impedance measurements are collected for a particular type of lesion, tissue, or other anatomy, one or more of the collected impedance measurements correspond to another anatomy. Is also good. For example, probes used to collect impedance measurements for clots or other vascular lesions may not only come into contact with the clot alone, but may also be proximal to the clot, such as a vessel wall. May come in contact with other tissues. These impedance measurements on other tissues or anatomy can interfere with proper clot analysis. We have used one or more machines to distinguish between impedance measurements related to the anatomy of interest and impedance measurements that reflect other anatomy or reflect errors in data collection. Recognize and understand the benefits of training a model using learning techniques. If the system is trained to recognize features of a lesion (or other anatomy of interest) in a part of the animal's body, it can be found in that part of the body or One or more other anatomy that would have been may be identified. Impedance measurements can be collected for these other anatomy. Once collected, this impedance measurement can be used in parallel with other impedance measurements on the lesion (or other anatomy of interest) to measure impedance on the lesion (or other anatomy of interest) or other structure. The model can be trained to distinguish between. After training, the model can be used to filter the input impedance measurements and prevent or mitigate the possibility of impedance measurements on other anatomy that interfere with the analysis of the impedance measurements on the lesion.

  It should be understood that not all lesions are formed in the vessel, and that some embodiments may work with lesions in areas of the body other than the vessel. For example, some cancerous cells may have formed in other parts of the body of an animal (eg, a human). Some embodiments described herein relate to the diagnosis and / or treatment of lesions, such as cancerous cells, that are not normally found in ducts. However, it should be understood that some cancerous cells can be found in the ducts and that other embodiments described herein relate to the diagnosis and / or treatment of such cancerous cells. .

  Also, although some of the examples described below relate to lesions, embodiments are not limited to working with lesions, and any target biological material with any suitable biomaterial composition. It should also be understood that the structure can work.

DETAILED DESCRIPTION OF THE TECHNOLOGY FIG. 1 provides a context for discussion of exemplary components of a medical device operating according to some embodiments described herein, 5 is a flowchart of a process that a clinician may follow. FIGS. 2-3 illustrate examples of medical devices, and the other figures below detail other components of the device and the manner in which the device can operate.

Process 100 may be used to diagnose and / or treat a lesion in an animal subject. The animal can be a human or non-human animal, including, for example, a human or non-human mammal. The lesion may be a lesion in a duct, for example, a blood vessel such as a vein or artery of an animal. Vascular lesions may completely or partially obstruct the tube. Embodiments described herein include:
Blood clots (including red blood cells, white blood cells, fibrin, thrombus, emboli, and / or platelets) formed at the site of the lesion in the vasculature or elsewhere in the body and fixed at the site of the lesion;
Tumors from the vessel wall to the center of the vessel, eg, tumors of scar tissue or other tumors after damage to endothelial cells at the site of the lesion;
Tissues that are not anatomically "normal" or "healthy" with respect to the vessel at the site, but otherwise extend from the vessel wall to the center of the vessel (eg, smooth muscle cells, elastic fibers, outer elastic lamina, internal elastic lamina) , Loose connective tissue, and / or endothelial cells);
Cholesterol, calcium, fatty substances, cellular waste products, fibrin, and / or other substances (eg, in the case of vascular disease, the blood vessels of the animal) that may be found in body fluids flowing through the animal's ducts Accumulation of plaque-like substances at the site of the lesion, including the accumulation of substances found within
It can operate on lesions of different characteristics, such as cancerous cells found in ducts such as metastases and / or lymphomas; and / or any other tissues and / or biological material that can cause lesions in animal ducts.

  Lesions of different characteristics can form outside the vessel. These lesions include cancerous cells, eg, cell tumors, myeloma, leukemia, lymphoma, melanoma, neoplasm, mixed and / or sarcoma.

  In some embodiments, a histology of the lesion (eg, which of the above-described biological materials the lesion has) is analyzed by a plurality of impedance spectra for the lesion, the composition of the lesion (the composition being present in the lesion). (Which may represent biological material). Identification of such a biological tissue can be performed by determining the relative amount of the tissue and / or cells present in the lesion and / or the plaque-like substance present in the lesion and / or the tissue, cells or plaque-like substance present in the lesion Identification may be included. In some embodiments, identifying the biological material present in the lesion may include, for example, identifying the status of each biological material with respect to the tissue / cell, regardless of whether the tissue / cell is healthy or unhealthy. . An unhealthy state of a cell can include, for example, whether the cell is in an inflammatory state, a disease state, a cancerous state, or other abnormal state.

  It should be understood that embodiments are not limited to operating with any particular shape or composition of lesions in the subject's anatomy, or with any particular location of lesions. As mentioned above, for simplicity of explanation, various examples are provided below in which the vessel is an animal vasculature.

  Prior to initiating the process 100 of FIG. 1, the subject may have exhibited symptoms of a vasculature lesion. For example, a clinician may make a tentative determination whether there is a lesion and a potential location of the lesion using an imaging technique such as angiography. Based on the provisional determination of the symptoms and location of the lesion, the clinician may choose to insert an invasive device into the subject's vasculature to further diagnose and / or treat the lesion. The clinician may be, for example, a physician (eg, a physician or surgeon) or another medical professional, such as a nurse or medical technician operating a medical device (perhaps under the supervision of a physician). It may be. In some embodiments, the clinician may be located in the same room as the subject, including next to the subject, and in other embodiments, the clinician may be remote from the subject. At a location (eg, in a different room in the same building as the patient, or geographically separated from the patient) via one or more wired and / or wireless networks, including the Internet or other wide area network (WAN) Operating a user interface for controlling the medical device.

  The process 100 begins at block 102, where a clinician inserts an invasive probe into the vasculature of the subject. The invasive probe inserted by the clinician at block 102 may be located at the distal end of a guidewire for a medical device and may be shaped, sized, or configured for insertion into the vasculature. Further, at block 102, the clinician may deliver the invasive probe through the subject's vasculature until the invasive probe is positioned proximal to the lesion. To do this, the clinician may use imaging techniques, such as angiography techniques, to monitor the position of the invasive probe in the subject. Insertion and delivery of the invasive probe at block 102 may be performed using any suitable technique for insertion of the device into the vasculature, including using known techniques, and embodiments Not limited.

  At block 104, the clinician activates the invasive probe to determine one or more characteristics of the lesion. The characteristics may include a phenotype and / or genotype of the anatomy, such as a lesion, including a property that distinguishes between anatomies or distinguishes between phenotypes of the anatomy. A feature may be a property that affects the treatment of the lesion (or other anatomy), and a lesion having this property may be treated from a lesion that does not have this property, and may have different values for this property. Lesions with a may be treated differently. Such properties may be histological in relation to the anatomy of the lesion and / or how the lesion is located in or interacts with the animal's body It can be anatomically related to crab. Thus, the features may explain the lesion. Exemplary features include the location of the lesion, the size (eg, length) of the lesion, the composition of the lesion, or other features discussed in detail below. To determine the characteristics, one or more sensors of the invasive probe may be used to detect the tissue of the lesion and / or other biological material and / or another tissue at the site of the lesion, such as healthy tissue located near the lesion. One or more measurements of a substance may be made.

  In some embodiments, determining one or more characteristics of the lesion may include identifying the composition of the lesion, for example, by identifying the amount of different types of cells or tissues present in the lesion. . As an example, it can be identified that the investigated pathology is composed of 50% red blood cells, 30% fibrin, and 20% platelets.

  Examples of sensors and measurements are described in detail below. To activate the invasive probe at block 104, the clinician contacts the lesion with one or more sensors of the invasive probe and / or activates a medical device user interface to detect characteristics of the lesion. In order to use the sensor, an invasive probe may be activated.

  At block 106, the clinician activates the medical device to create and output a treatment recommendation for the lesion based on the determined lesion characteristics. As discussed in detail below, treatment recommendations generated by the medical device based on the characteristics of the lesion may indicate how to treat the lesion, such as which treatment device to use to treat the lesion (eg, And whether to use a suction catheter or stent-type thrombus collector when the material of the lesion is to be removed from the subject) and / or how to use the treatment device (how fast the stent-type thrombus collector is withdrawn). May contain recommendations. As also detailed below, the medical device compares, for example, a condition associated with each of the plurality of different treatment options to a feature of the lesion, and the feature of the lesion satisfies a corresponding condition for the treatment option. In some cases, treatment recommendations may be created based on various analyzes, such as by outputting treatment option recommendations. The output by the medical device may be via any suitable form of user interaction, including visual, audible, and / or tactile feedback to the clinician via a user interface. In some embodiments, the medical device may automatically analyze the characteristics of the lesion determined at block 104 and create / output a treatment recommendation at block 106 without further user intervention. In other embodiments, the clinician may activate the medical device user interface to request the creation / output of an analysis and / or treatment recommendation.

  At block 108, the clinician reviews the medical device treatment recommendations, selects a treatment option, and at block 110 treats the lesion using the selected treatment option.

  In some embodiments, the selected treatment option may include the insertion of additional invasive medical components into the vasculature of the subject. If the invasive probe inserted at block 102 was a component of a guidewire, for example, additional treatment devices may be inserted along the guidewire. As a specific example of such a case, if the medical device recommends complete or partial removal of the lesion using a stented thrombectomy, the stented thrombus can be inserted into the vasculature. As another example, if the medical device instead recommends removal with a suction catheter, the clinician may insert the suction catheter into the vasculature. As a further example, if the medical device recommends implantation of a stent, a stent implanter may be inserted into the vasculature.

  In other embodiments, the treatment may not require the insertion of another device. For example, the invasive probe inserted at block 102 may not be a component of the guidewire, but may instead be a component of a treatment device, such as a stent thrombus collector. In such a case, the treatment of block 110 may be performed using the treatment device inserted at block 102. For example, if the invasive probe inserted at block 102 is a component of a stent thrombectomy, the treatment recommendations at block 106 may indicate the amount to expand the stent, the amount of time to wait for the clot to coalesce with the stent. It may be specific to the method of operating the stent thrombus collector, such as the amount and / or the force or speed for collecting the stent and clot. In such an embodiment, at block 110, the clinician may treat the lesion by activating the stent thrombus collector as recommended by the medical device at block 106.

  In some embodiments, a treatment recommendation can be made that includes treating the subject in a manner that does not treat the lesion, but rather leaves the lesion untreated. For example, some types of vascular lesions may be difficult to treat effectively, preventing reopening of blood vessels obstructed by the lesion. For example, lesions that reflect intracranial atherosclerotic disease (ICAD lesions) may be difficult to rule out certain treatments currently available. Thus, in some embodiments, if an ICAD lesion is detected, the medical device may make a treatment recommendation not to treat the lesion until an effective ICAD treatment is available. ICAD lesions can be identified by their composition, particularly the location of the biological material within the lesion. For example, a medical device may, via the techniques described herein, cause a vascular lesion to develop a blood clot on the luminal side of an invasive probe, and atherosclerotic plaque on the opposite side of the probe lumen. If the medical device determines that the lesion is an ICAD lesion, the medical device may determine that the lesion is an ICAD lesion (eg, lipid or calcified components, smooth muscle cells, absence of endothelium).

  After the lesion has been treated at block 110, process 100 ends. Additional actions after treatment of a lesion that may be performed in some embodiments are described below.

Examples of Medical Devices As described above, FIG. 1 illustrates that a medical device may operate according to some embodiments described herein to diagnose and / or treat a lesion in an animal's vasculature. A general discussion of the method was provided. FIGS. 2-3 provide examples of some embodiments of medical devices that include invasive probes that can be inserted into the vasculature as part of the diagnosis and / or treatment.

  FIG. 2 illustrates a medical device 200 that a clinician 202 can operate to diagnose and / or treat a condition of a subject 204. The condition of the animal 204 (eg, a human) can be a vasculature lesion 204A, exemplified in the example of FIG. 2 as a lesion in a blood vessel in the human head that can cause ischemic stroke. As described above, lesion 204A can be a blood clot, plaque-like substance accumulation, excessive proliferation of smooth muscle tissue, and / or other vascular lesions.

  The medical device 200 illustrated in FIG. 2 includes a guidewire 206, a handle 208, and an invasive probe 210. At least a portion of the invasive probe 210 and the guidewire 206 may be inserted into the vasculature of the subject 204 until the invasive probe 210 is located proximal to the lesion 204A. Thus, the invasive probe 210 may be shaped and configured for insertion into the vasculature (or other vessel). In some embodiments, the invasive probe 210 is a guidewire that is about 300 micrometers, or a microcatheter that is about 3-10 French in diameter, or another device that has a diameter suitable for insertion into a vessel of an animal. Attached to. Such a device, in some such embodiments, may be about 1 or 2 meters in length, where the invasive probe 210 may have a guidewire, for example, within 5 cm of the end of the device. / Located at one end of the device.

  The invasive probe 210 inserted into the subject 204 may include one or more sensors 212 and a measurement unit 214. In some embodiments, sensor 212 may measure one or more electrical characteristics of lesion 204A, which may measure one or more electrical characteristics of tissue and / or biological material of lesion 204A. Includes measuring by measuring. The measurement unit 214 may receive data generated by the sensor 212, and in some embodiments, one or more of the data applied to the lesion 204A as part of measuring one or more electrical features. One or more electrical signals may be created.

  Examples of the sensor 212 are described in detail below. As a specific example, the sensor 212 may be an impedance sensor, and the measurement unit 214 may cause the sensor 212 to perform electrical impedance spectroscopy (EIS) of the lesion 204A. For example, the measurement unit 214 is a particular frequency selected to distinguish between different tissues and / or different biological materials to help identify the composition of the lesion 204A, as discussed in detail below. It may include one or more oscillators for generating an electrical signal of one or more frequencies (and configured by the oscillator of the measurement section 214). In embodiments configured to test tissue / substances using multiple frequencies, the measurement unit 214 may be configured such that one oscillator is specific to each frequency tested and generates a signal at this frequency. It may include a plurality of oscillators configured.

  In some embodiments, where the measurement section 214 generates an electrical signal to apply to the lesion 204A, the measurement section 214 is included in the invasive probe 210 and inserted into the vasculature of the subject 204. May be preferred. This may place the measurement section 214 very close to the sensor 212 and the lesion 204A, and may limit the noise of the electrical signal applied to the lesion 204A. For example, when the measuring unit 214 is disposed on the handle 208, the electric signal generated by the measuring unit 214 is output by the invasive probe 210 after moving from one end of the guide wire 206 to the other and applied to the lesion 204A. You. As the signal travels across the guidewire 206, electrical noise may affect the quality of the signal. By placing the measurement section 214 within the invasive probe 210, the noise of this signal can be limited. When the measurement unit 214 is disposed in the invasive probe 210, it may be disposed in the lumen of the invasive probe 210, or may be disposed on the surface (internal or external) of the invasive probe 210. Or it may be embedded in a film attached to the surface (internal or external) of the invasive probe 210.

  The measurement unit 214 may be configured as an application specific integrated circuit (ASIC) in some embodiments. In some such embodiments, the ASIC may be manufactured using a packaging process that reduces the silicon substrate layer. For example, during fabrication, an integrated circuit may be fabricated with an "active" silicon layer that includes functional components on top of a silicon substrate layer that does not contain active elements. This substrate layer may be the lowest layer in the stack of layers, and optionally the thickest layer. Conventionally, the substrate layer remains unprocessed after fabrication to provide structural stability to the integrated circuit. In some embodiments, the measurement circuit 214 may be manufactured using a process that includes removing the silicon substrate layer after manufacturing the active layer and before packaging. The manufacturing process may include removing the substrate from the bottom surface of the wafer, which may be the side opposite the side on which the active device was manufactured. In some embodiments, all of the silicon substrate may be removed. In other embodiments, substantially all of the silicon substrate may be removed, where "substantially" removing removes the active layer components without leaving the silicon substrate solely for structural support. Leaving only enough silicon substrate to ensure proper electrical function of the substrate. After removing the silicon substrate, the integrated circuit may be wrapped in a packaging material.

  In some embodiments, placing the measurement portion 214 in close proximity to the sensor 212 and lesion 204A may reduce signal attenuation by limiting the distance traveled by the electrical signal. This reduction in signal attenuation can be particularly significant at higher frequencies because wires tend to exhibit a lower frequency response. By increasing the cutoff frequency of the electrical path between the signal source and the lesion by reducing the distance traveled by the signal, the range of frequencies available for diagnosis or therapy may be increased. As a result, properties for distinguishing tissue or cell types can be significantly enhanced. Placing the measurement section 214 in close proximity to the sensor 212 and lesion 204A may result in a cutoff frequency of up to 1 MHz in some embodiments, up to 10 MHz in other embodiments, and up to 10 MHz in yet other embodiments. It can be increased to 25 MHz. For comparison, when the measurement unit 214 is disposed on the handle 208, the cutoff frequency may be limited to less than 500 kHz.

It should be understood that embodiments are not limited to whether sensor 212 is an EIS sensor or is driven to perform an EIS process. In some embodiments, sensor 212 may be or include one or more electrical, mechanical, optical, biological, or chemical sensors. Specific examples of such sensors include an inductance sensor, a capacitance sensor, an impedance sensor, an EIS sensor, an electrical impedance tomography (EIT) sensor, a pressure sensor, a flow sensor, a shear stress sensor, a mechanical stress sensor, and a deformation sensor. , temperature sensor, pH sensor, a chemical composition sensor, (e.g., O ₂ ions, biomarkers or other compositions), the acceleration sensor, and motion sensors, and the like. These sensors may include known commercially available sensors.

  In some embodiments, the measurement portion 214 included in the invasive device 210 drives the sensor 212 and / or processes the results from the sensor to create data and handles along the guidewire 206 208 may be configured to send data back. This may be the case, for example, for an embodiment where a treatment advisory is created by the medical device 200. Data characterizing the lesion 204A may be transmitted along the end of the guidewire 206. In order to limit the effects of noise during such transmissions, in some embodiments, measurement unit 214 may communicate over a communication channel (eg, one or more wires) that runs over guidewire 206. An analog-to-digital converter (ADC) or other components may be included to create digital data for transmission.

  In accordance with embodiments described herein, clinician 202 may treat lesion 204A with one or more treatment recommendations generated by medical device 200. Although not illustrated in FIG. 2, the medical device 200 may include a controller for generating and outputting the treatment recommendation for treating the lesion 204A. The controller may, in some embodiments, be implemented as a lesion analysis facility, implemented as executable code executed by at least one processor of the medical device 200. The lesion analysis facility may analyze characteristics of the lesion 204A determined by the medical device 200 (eg, by the invasive probe 210) in association with information set for one or more treatment recommendations. As a specific example, as discussed in detail below, the lesion analysis facility may compare the characteristics of the lesion 204A with conditions associated with various treatment recommendations, and the conditions of the treatment recommendations with the characteristics may be used. If is satisfied, the treatment recommendation can be output.

  In some embodiments, a processor for executing the lesion analysis facility and a storage medium (e.g., a memory) storing the lesion analysis facility, and information set regarding the treatment recommendation are disposed in the handle 208. Is also good. Thus, a lesion analysis facility executed by the processor at handle 208 may receive data representative of one or more features of lesion 204A from measurement unit 214 via the communication channel of guidewire 206.

  In other embodiments, however, the processor configured to execute the lesion analysis facility and the storage medium (eg, memory) storing the lesion analysis facility, and the information set regarding the treatment recommendations are located, for example, on another computer device. By doing so, it may be arranged separately from the guide wire 206 and the handle 208. This computing device may be located proximal to guidewire 206 and handle 208, for example, by being located in the same room. Alternatively, the computing device may be spaced apart from guidewire 206 and handle 208, for example, by being located in a different room of the same building, or may be geographically separated from guidewire 206 and handle 208. . In embodiments where the processor / media is spaced apart from the guidewire 206 and the handle 208, the computing device may be a direct wire from the handle 208 to the computing device, a wireless personal area area between the handle 208 and the computing device. A network (WPAN), a wireless local area network (WLAN) between the handle 208 and the computing device, a wireless wide area network (WWAN) between the handle 208 and the computing device, and / or one or more wired and And / or via a wireless communication network, data representative of one or more features of lesion 204A may be received. Thus, in some embodiments, handle 208 may include one or more network adapters for communicating over one or more networks.

  If the treatment advisory is created by the medical device 200, the treatment advisory may be output by the medical device for presentation to the clinician 202 and / or any other user. This output may be output to another device and / or one or more displays, such as display 216, or other forms of user interface, via one or more networks. In the example of FIG. 2, the lesion analysis facility may execute on a processor located within handle 208 and may generate a treatment recommendation, which may be displayed on display 216 for presentation to clinician 202. To the output of the handle 208 via the wireless network adapter. Embodiments are not limited to this form, and other forms of user interface may be used. Any suitable visual, audio, and / or tactile feedback may be used. For example, if a treatment recommendation is made during the removal of a lesion using either a suction catheter or a stent-type thrombus collector, the handle 208 may include a light emitting diode (LED) or other visual indicator for each option. Components may be included and treatment recommendations may be presented by brightening the appropriate LED. As another example, if the treatment recommendation relates to a method of activating a stent clot collector, and particularly to a recommendation of when to initiate a withdrawal after a waiting time, the signal to initiate withdrawal may include a handle 208. May be output using a tactile signal provided via a vibration unit incorporated into the device. One skilled in the art will appreciate that, similar to the computing device described above, the components of the user interface may be located within the handle 208 or may be spaced (or even far away) from the handle 208. Understand that.

  Power may be provided to the invasive probe 210 via a power cable extending along the end of the guidewire 206. The power cable may be connected to a power supply of the handle 208, which may be a battery, an energy harvester, a connection to a grid power supply, or other source of energy, but in embodiments Is not limited to this mode.

  In some embodiments, handle 208 may include one or more sensors not illustrated in FIG. Sensors integrated into handle 208 may monitor the operation of medical device 200 to convey how clinician 202 has performed treatment. For example, an accelerometer or other motion sensor may be located on handle 208 to detect movement of handle 208 that coordinates movement of guidewire 206 and invasive probe 210. For example, by monitoring the accelerometer, the clinician 202 may have performed multiple treatments (e.g., multiple passes using a suction catheter or stent thrombectomy) to remove the lesion, or may have performed one pass. A determination can be made as to whether the lesion was recovered alone.

  In some embodiments, handle 208 may be removable from guidewire 206 and may be reusable with each actuation. Thus, the invasive probe 210 and / or the guidewire 206 may be configured to be non-reusable, and instead may be configured to be discarded for hygienic reasons, and the handle 208 Is removably attached to guidewire 206 and may be configured to be reused with other guidewire 206 and invasive probe 210. For example, the guidewire 206 and the handle 208 may have a complementary interface such that the handle 208 can connect with the guidewire 206 and interface with components of the guidewire 206 (eg, communication channels, power cables) and invasive probes. May be provided.

  The clinician 202 can operate the medical device 200 via a medical device user interface that includes the display 216 and can be at least partially disposed within the handle 208. For example, handle 208 allows clinician 202 to move guidewire 206 and invasive probe 210 forward or backward through the vasculature, and / or effect actuation of invasive probe 210.

  Operation of the invasive probe 210 may depend on the components of the invasive probe 210. For example, invasive probe 210 may include a sensor 212 for detecting one or more features of lesion 204A. In addition, the invasive probe 210 may, for example, activate one or more sensors to apply an electrical signal to the lesion 204A and make one or more measurements of the lesion 204A during and / or after the application of the electrical signal. And a measurement unit 214 for activating the sensor to detect one or more features. In some embodiments, the invasive probe 210 may include one or more components for treating the lesion 204A, including treatment by implanting a stent and / or removing the lesion 204A. . The components for lesion removal may include components related to any technique suitable for removing lesions, but embodiments are not limited to this form. In some embodiments, for example, the invasive probe 210 may aspirate the lesion into a stent-type thrombectomy component (eg, a balloon) and / or a catheter for performing stent-based retrieval of the lesion. A suction catheter component to perform Further, invasive probe 210 may include other sensors not shown in FIG. 2, such as, for example, optical coherence tomography (OCT) sensors.

  The medical device user interface may be integrated into all or part of the handle 208, thus allowing the clinician 202 to perform many different operations with the invasive probe 210. For example, the user interface of the handle 208 allows the clinician 202 to apply electrical signals to the sensor 212 and the measurement unit 214 and / or to make a measurement of the lesion 204A and / or treat the lesion 204A. For one or more treatments.

  While an example has been described in which the medical device 200 may include a treatment component to perform one or more actions to treat the lesion 204A, it should be understood that embodiments are not limited thereto. In some embodiments, the medical device 200 is positioned proximal to the lesion 204A, with a guidewire for an additional treatment device inserted along the guidewire to treat the lesion 204A. possible. For example, after insertion of invasive probe 210 and guidewire 206, clinician 202 may insert another device along the length of the guidewire, or remove guidewire 206 and invasive probe 210. , Then a new device can be inserted. The newly inserted device can be, for example, a stent implanter, a suction catheter, a stent-type thrombus collector, or other device for treating the lesion 204A. In some embodiments where an additional device is inserted, the handle 208 may be configured such that the additional device and the handle 208 may have a compatible interface, and the additional It can be compatible with additional devices so that the device can operate.

  Further, while examples have been provided in which the clinician 202 manually operates the medical device 200 according to treatment recommendations, embodiments are not limited thereto. In an alternative embodiment, the medical device 200 may automatically treat the lesion based on input from the sensor 212. For example, as is evident from the concise discussion above and the following detailed discussion, medical device 200 may formulate a treatment recommendation for a method of treating lesion 204A. In some embodiments, the medical device 200 is a suction catheter for treating the lesion 204A according to treatment recommendations without user intervention (but, in some embodiments, under the supervision of the clinician 202). Insert and / or activate a stent-type thrombus collector, stent implantation device, or other device according to treatment recommendations.

  It should be understood that embodiments are not limited to operation with medical devices that are invasive or that include invasive components inserted into the body of an animal. For example, non-invasive probes may operate using the selected frequencies or characteristics described herein, or include using models trained as described herein. It may have a measurement and / or sensor (such as an EIS sensor) that operates as described herein. Such non-invasive devices can be used, for example, in diagnosing and / or treating skin lesions.

  Also, it is noted that the techniques described herein are not limited to use with insertable devices, such as guidewires or other tools, which can be inserted and then removed, but may also be used with implantable devices. Should be understood. For example, a measurement portion and sensor of the type described herein can be used with a stent, for example, a stent where the sensor is located directly on the stent. In this manner, monitoring of the tissue in the area where the stent is located may be performed immediately after and after the stent is deployed. The sensor may detect one or more characteristics (eg, composition) of the tissue in the area where the stent is located. The sensed features may be used to infer characteristics of one or more anatomical structures with which the stent has contacted and to make decisions regarding the one or more anatomical structures. For example, the system may be used to determine whether the tissue that the stent is in contact with is healthy, or whether scar tissue or other unhealthy tissue is forming, or whether an occlusion has been formed. , Can be used.

  FIG. 3 illustrates an example of an invasive probe 210 in which some embodiments may operate. The example invasive probe 210 of FIG. 3 includes a mesh 300 that is configured similar to a stent. Invasive probe 210 may, in some embodiments, be operable as a stent-type thrombus collector. In other embodiments, the invasive probe 210 is not operable as a stented thrombectomy, but may have a sensor to detect lesion features with better accuracy than perhaps using only one sensor. May include a mesh 300 or another structure to provide multiple points of contact between the and the lesion.

  Thus, it is understood that in some embodiments (not the embodiment of FIG. 3), the invasive probe 210 may include only one sensor, which may be located at the distal end of the invasive probe 210, for example. Should. Such a sensor may be implemented as two electrodes, one of which may apply an electrical signal to the lesion, one of which may receive the applied signal. Based on a comparison of the received signal and the applied signal, one or more decisions may be made, as discussed in detail below.

  However, the present inventors have recognized and appreciated that by including additional sensors in the invasive probe 210, it may be possible to determine more detailed information. For example, including an additional sensor in the invasive probe 210 may be able to create information about the composition of the lesion with greater accuracy than only one sensor. Such additional sensors may, for example, determine the impedance spectrum at each of a plurality of locations along the invasive probe, such that possibly different impedance spectra may be determined at different locations of the same lesion. Can be This may include, for example, determining an impedance spectrum using each sensor. Each impedance spectrum in this case would be the impedance spectrum of the biological material of the lesion that the sensor (with its two electrodes) contacts. Some lesions may include multiple different biological materials (eg, different tissues or cells, or different plaque-like materials). If each sensor of the invasive probe contacts a different biological material, each sensor may determine a different impedance spectrum for each different biological material. However, this may be the case where, in some lesions, two or more sensors of the invasive probe may come into contact with the same biological material, in which case the same or substantially the same impedance spectrum will be obtained. Can create. Thus, in some embodiments, the invasive probe may operate each sensor to create an impedance spectrum for the biological material of the lesion. Creating an impedance spectrum for each of the plurality of biological materials of the lesion (ie, a plurality of impedance spectra for each lesion) is in contrast to determining one impedance for the lesion as a whole. Techniques for determining the composition of a lesion using multiple sensors, including through performing an EIS, are discussed below.

  Thus, FIG. 3 illustrates an example of an invasive probe 210 having a plurality of sensors located along the outer and / or inner surface of the probe 210. Sensors 302 (sensors 302A, 302B, 302C, 302D, generically or collectively referred to herein as sensors 302) may be disposed along structure 300. In some embodiments, each sensor may be or include one or more electrodes for applying an electrical signal and / or detecting the applied electrical signal.

  In some embodiments, not illustrated in FIG. 3, the invasive probe 210 can include a balloon to expand the structure 300 outward when inflated and make good contact with the lesion. During use, for example, the structure 300 may be a sensor located at the distal end of the structure 300, for example, after inflating the structure 300 using a balloon until the sensor 302 detects contact at multiple points. Can be inserted into the lesion, in whole or in part, until they detect that they pass through the far side of the lesion. Inflation of the structure 300 may be controlled by a controller of the invasive probe 210 (eg, the measurement unit 304), or by a lesion analysis facility located elsewhere on the medical device, and / or by a user of the medical device. It may be controlled by a clinician via an interface.

  In some embodiments, measurement unit 304 may include one or more electrical signals, including, for example, by generating one or more electrical signals for application to a lesion and analyzing the data generated by sensor 302. The sensor 302 may be activated to make a measurement. Analysis of the data generated by the sensor 302 may include performing an analog-to-digital conversion of data transmitted along the guidewire to the outside of the patient, such as a lesion analysis facility or user interface described above.

While examples have been provided in which sensor 302 is an electrical sensor, it should be understood that embodiments are not limited thereto. For example, the sensor 302 may be or include one or more electrical, mechanical, optical, biological, or chemical sensors. Specific examples of such sensors include an inductance sensor, a capacitance sensor, an impedance sensor, an EIS sensor, an electrical impedance tomography (EIT) sensor, a pressure sensor, a flow sensor, a shear stress sensor, a mechanical stress sensor, and a deformation sensor. , temperature sensor, pH sensor, a chemical composition sensor, (e.g., O ₂ ions, biomarkers or other compositions), the acceleration sensor, and motion sensors, and the like.

Examples of Sensors and Sensing Techniques As described above, in some embodiments, the measurement portion of an invasive probe may operate the sensor of the invasive probe to perform electrical impedance spectroscopy (EIS). 4 to 11 describe examples of how such sensors and measuring units can be arranged and describe examples of techniques relating to the operation of the sensors and measuring units. However, it should be understood that embodiments are not limited to operation by the example for the EIS described in this section.

  The techniques described in this section with respect to FIGS. 4-11 allow the differentiation of tissue and / or biological material of vascular lesions in animals, including mammals such as humans. As used herein, "distinguish" refers to, for example, one or more types of cells of a lesion (eg, red blood cells and / or white blood cells, or endothelial cells of a different type or state), and / or one or more types of lesions. It should be understood that determining the other biological material (eg, a plaque-like material such as cholesterol) implies the potential offered by this method to distinguish between lesions of different composition. More generally, the distinction, possibly made by the techniques described in this section, involves determining at least one item of information related to the pathology tested. Examples of information that can be determined via these techniques are provided below.

  The cell differentiation method 10 schematically illustrated in FIG. 4 includes a first step 12 of determining the frequency spectrum of the impedance of the lesion to be tested.

  As used herein, a spectrum is to be understood as meaning a set of impedance pair values of a lesion, the latter of which can be a complex number and the corresponding frequency value. Thus, the spectrum may be discrete and may include only a finite number of pairs. These pairs may be separated, in particular, by a few Hz, even by tens of Hz, and even by hundreds of Hz. However, in other embodiments, the spectrum determined in this step is continuous, pseudo-continuous, or discrete over the frequency band. Pseudo-continuous should be understood to mean that the spectrum is determined at a continuous frequency, separated by less than 100 Hz, preferably less than 10 Hz, more preferably less than 1 Hz. The frequency band in which the impedance of the tissue is determined extends, for example, from 10 kHz, preferably from 100 kHz. In fact, at low frequencies, the lesion tissue / substance membrane acts as an electrical insulator, so that the impedance is very high and, in particular, varies little. Furthermore, the frequency band in which the impedance of the tissue / material of the lesion is determined extends, for example, up to 100 MHz, preferably to 1 MHz. In fact, at high frequencies, the tissue / material walls that make up the lesion become permeable from an electrical point of view. Thus, the measured impedance no longer represents the biological material. This spectrum may be a real and / or imaginary part and / or an absolute value and / or a phase frequency spectrum of the complex impedance of the lesion.

  The first step 12 of determining the frequency spectrum of the impedance of the lesion may be performed as described hereinafter in particular in connection with FIG.

  First, during step 14, two, preferably three, and even more preferably four electrodes are placed in contact with the lesion to be tested. These electrodes are connected to an alternator. Measurement with four electrodes is preferred because it allows the use of two electrodes to pass current through the lesion to be tested and the potential difference between the other two electrodes. This makes it possible to improve the accuracy of the measurement. Next, during step 16, an alternating current is applied between the electrodes in contact with the lesion. Next, during step 18, a corresponding voltage is measured at the end of the electrode for a different frequency by changing the frequency of the applied current. Finally, during step 20, the ratio between the measured voltage and the applied current was calculated at each frequency where the measurements were taken. This ratio provides the impedance of the examined lesion as a function of the measured frequency. The calculated ratio makes it possible to define the frequency spectrum of the impedance of the lesion.

  If the spectrum is continuous or pseudo-continuous, this may be represented in the form of a curve providing the absolute value of the impedance of the lesion as a function of frequency in this particular case, as illustrated in FIG. The latter is plotted on a log scale. Note that a log scale is used here on the x-axis.

  In step 22 of the discriminant method 10 of FIG. 4, a different model of the impedance of the lesion is selected, that is, a different electric circuit capable of modeling the lesion. Here, a model that includes the constant phase element but does not include the capacitance is selected. In fact, it has been found that constant phase element models model lesion events more realistically than capacitance.

（式中、
ｊは、−１の平方根（ｊ^２＝−１）であり；
ωは、電流の特定のパルスであり（ω＝２πｆ（ｆは電流の周波数である））；
Ｑ_０は、定位相要素の実パラメータであり；
αは、０〜１の間にある定位相要素の別の実パラメータであり、これにより、定位相要素の相φ_ＣＰＥは、−απ／２に相当する）
を有する。 The constant phase element (or CPE) is in the form of an impedance Z _CPE :

(Where
j is the square root of −1 (j ² = −1);
ω is a particular pulse of current (ω = 2πf, where f is the frequency of the current);
Q ₀ is the real parameter of the constant phase element;
α is another real parameter of the constant phase element between 0 and 1, whereby the phase φ _CPE of the constant phase element corresponds to −απ / 2)
Have

  In the following description, a constant phase element whose impedance is provided by equation [1] above has been selected as an example.

  The model of the impedance of the lesion may be selected from among those described herein below with respect to FIGS. 7-10, among others. Obviously, the simpler the model, the simpler the calculation. However, the complex model may better correlate with the impedance spectrum obtained from the measurement, and may thus give more accurate results.

（式中、
Ｚ_ｔｏｔは、病変を表す第１のモデル２４の総インピーダンスであり；
Ｒ１およびＲ２は、第１の抵抗２６および第２の抵抗３２の抵抗値である）
の形態である。 In the first model 24 illustrated in FIG. 7, the impedance of the lesion is modeled by a first resistor 26 mounted in series with a parallel connection 28 of a constant phase element 30 and a second resistor 32.
In this case, the total resistance Z _{tot of the} lesion is

(Where
Z _tot is the total impedance of the first model 24 representing the lesion;
R1 and R2 are the resistance values of the first resistor 26 and the second resistor 32)
It is a form of.

  Such a model describes the lesion covering the measuring electrode particularly well, such as a set of individual parallel mountings, where each individual mounting is in series with the parallel mounting of individual resistances and individual capacitances. Are composed of individual resistors. Such an attachment makes it possible to model the time constant of the distribution of the time constant over the entire surface of the measuring electrode by different circuits in parallel, whose parameters may be different. Here, each of these parallel circuits represents a different tissue / substance of the lesion. Thus, the fact that the tissue / material of the lesion may exhibit different electrical properties, especially different resistances and / or capacitances, is modeled.

（式中、
βは、この第２の定位相要素の定位相が−βπ／２に相当するような、０〜１の間にある実パラメータであり；
Ｑ_１は、定位相要素の実パラメータである）
の形態であるように選択され得る。 The second model 34 illustrated in FIG. 8A complements the model 24 of FIG. 7 by the mounting of a second constant phase element 36 in series. The impedance Z _{CPE, 2} of this second constant phase element 36 is likewise

(Where
β is a real parameter between 0 and 1 such that the constant phase of this second constant phase element corresponds to -βπ / 2;
Q ₁ is the real parameter of the constant phase element)
May be selected to be in the form of

により提供される。 Therefore, the total impedance Z _{tot of the} lesion according to the second model 34 is represented by the following equation:

Provided by

  A variant 34 'of the second model 34 is shown in FIG. 8B, with the addition of a capacitance C in parallel with the circuit of FIG. 8A for a good fit of the impedance curve at high frequencies. Different from the model.

により提供される。 The third model 38 illustrated in FIG. 9 corresponds to the third resistor 40 and the model of Figure 7 mounted in parallel of a resistor R _3. In this case, the total impedance Z _{tot of the} lesion is given by the formula:

Provided by

  Finally, a fourth exemplary model 42 is illustrated in FIG. This model 42 includes a first resistor 26 mounted in parallel with a series mounting of a constant phase element 30 and a second resistor 32 as illustrated.

により提供される。 The total impedance Z _tot of the lesion for this model 42 is given by the formula:

Provided by

  Next, a differentiating step determines, for each model selected in step 22, the impedance of the constant phase element 30 that optimizes the correlation between the model of the lesion impedance and the spectrum determined in step 12 (step 44). To continue.

  This step of optimizing the correlation between the model of the impedance of the lesion and the spectrum determined in step 12 may be performed by any optimization method known to those skilled in the art. By way of example, a least squares method may be implemented, which allows this step 44 to be implemented relatively simply in practice.

  In fact, other parameters of the different models, other than the constant phase element impedance parameters, are also determined during this step 44. These factors may also be useful for obtaining information about the pathology being examined and / or the tissue / material being composed.

  Next, an intermediate step 46 of the differentiator 10 may be provided. This step 46 involves determining the model that appears to best correlate with the measured spectrum of lesion impedance. This best model may be, for example, a model that minimizes the standard deviation in the measured spectrum. In the following description of the present specification, it is assumed that the model 24 is held as a model that best correlates to the measured spectrum of the impedance of the lesion.

  During step 48, the effective capacitance (or apparent capacitance) of the lesion is estimated from the constant phase element impedance parameters and the corresponding model.

  Theoretically, this effective capacitance represents the set of individual capacitances of the elements of the cell structure. Effective capacitance refers to the distributed local capacitance of the elements of the cellular structure. These elements of the cell structure are, among other things, the nucleus of the cell of the cell structure and other parts of the cell such as the Golgi, vesicles, mitochondria, lysosomes, and other elements that may play a role in membrane interactions. May be all or part of Also, the effective capacitance can be affected by cell geometry and the space between cells. Effective capacitance is a model that can represent some or all of the electrical membrane events of a lesion. This model allows the tissues / materials of the lesion to be relevantly distinguished.

  More substantially, the effective capacitance is determined by identifying the impedance of the lesion in a model that includes individual parallel attachments, where each individual parallel attachment is at least one individual resistance and 1 Includes two individual capacitances. Each mounting may in particular comprise and preferably consist of a first individual resistor in series with the parallel mounting of the second individual resistor and the individual capacitance. These individual attachments are aimed at modeling the events of each tissue / material of the lesion. Thus, the effective capacitance is the capacitance resulting from the presence of all individual capacitances at the lesion.

  For the model 24 (or 34 or 34 '), the determination of the effective capacitance can be performed in particular as follows. The impedance of model 24 with a constant phase element is compared to the impedance of an equivalent or identical model where the constant phase element has been replaced by an effective capacitance. Next, the calculation, or strictly speaking, the calculation of the effective capacitance, is based on the real part of the impedance of the same model in which the constant phase element is replaced by the effective capacitance and the model selected for the lesion with the constant phase element and / or This can be done by comparing the imaginary part and / or the phase and / or the absolute value.

を、式［３］から直接推定したモデル２４のアドミッタンスの式に導入することにより、以下の式［８］を得る。

ここから、実効キャパシタンスに関する式は、

の形態と推定され得る。 For the model 24 (or 34 or 34 '), for example, the time constant

Into the admittance equation of the model 24 directly estimated from the equation [3], the following equation [8] is obtained.

From this, the equation for effective capacitance is

Can be estimated.

When choosing another model of the impedance of a lesion with a constant phase element, it is possible to determine a corresponding equation for the effective capacitance. To do this, the impedances R ₁ , R ₂ of the model 24 or 34 or 34 ′, as a function of the parameters of the selected model, with respect to the model 24 or 34 or 34 ′ electrically corresponding to the model of the impedance of the lesion , Z _CPE , and Z _{CPE, 2} are adequately calculated. Thus, the effective capacitance can be calculated by replacing R ₁ , R ₂ , Z ₀ , and α with the corresponding values expressed as a function of the parameters of the selected model.

  Next, the cell differentiation method 10 continues with step 66 of inferring an item of information regarding the tissue / material of the lesion from the predetermined effective capacitance.

This inference can be made, inter alia, by comparing a pre-established value with the value of the effective capacitance determined in step 48. The pre-established values may in particular be obtained during tests carried out under known test conditions on tissues of known composition in known media. The pre-established values may be compiled into a database of values of effective capacitance, which summarizes the effective capacitance measured under different conditions of different types of cells and / or different cells and / or different test conditions. The value of the effective capacitance can be compared to a database of effective capacitances of cell types and conditions that are likely to be found in the measurements of the invention. In this comparison, the effective capacitance C _eff can be used together with other parameters. This comparison may not be an exact match, but may include determining whether the value of the effective capacitance is within or outside a predetermined range.

Thus, distinguishing the tissue / material of the lesion, ie the following information items:
Type of tissue and / or other biological material in the lesion;
Pathological composition, especially when the latter is composed of different types of biological material, or tissues / cells / other biological materials in different states;
When the lesion is composed of tissue, the type of cells contained in the tissue and / or the number of layers of cells present in the tissue;
If the lesion is composed of other biological material, such as plaque-like material, the type of substance contained in the lesion; and / or especially the cells are healthy, inflammatory, degenerated, especially 1 It is possible to determine at least one of the state of the cells contained in the diseased tissue when there is one or more cancerous cells in a degenerated state, when in an infected state.

  By way of example, FIG. 18 graphically depicts the effective capacitances 68, 70, 72, 74 determined in connection with tests performed by the previously described method.

Under the circumstances of the test, cells were cultured until confluence of the cells was obtained. In the case of the exemplary tests performed, two days of culture in a 37 ° C. incubator under 5% CO ₂ were required to obtain the tissue to be tested until confluence. Determination of the impedance spectrum of the different tissues to be tested was performed using an impedance spectroscopy system. The spectrum was determined to be between 1 kHz and 10 MHz by applying an alternating voltage which was rather low enough not to electrically excite the cells to be tested, but was assumed to be sufficient for accurate measurements. In the example of the test performed, the amplitude of the AC voltage of 20 mV was maintained.

  The effective capacitance 68 is the only static capacitance of the test medium. This test medium is a cell culture medium. The effective capacitance 70 is the capacitance of bovine aortic endothelial cells (BAEC). The effective capacitance 72 is the capacitance of bovine aortic smooth muscle cells (BAOSMC). Finally, the effective capacitance 74 is the capacitance of the blood platelets or thrombocyte. As the figure shows, the effective capacitances of different cell types are clearly different from each other, and it is possible to effectively distinguish between different cell types accurately without risk of confusion. It is possible.

  Thus, one advantage of the described discrimination method is that it allows for tissue / material distinction in a lesion in contact with an electrode from a simple measurement of the frequency spectrum of the impedance of the tested lesion. The results obtained are accurate. There is no need to perform normalization of the measured impedance, and no need to perform a reference measurement in the absence of any sample to be tested. Thus, the method can be performed without requiring prior sampling of the cells or cell structures to be tested, and in some embodiments, can be performed in vivo.

  It should be noted that in determining the effective capacitance, this single value is often sufficient to distinguish the diseased tissue / material. Also, the parameters of the selected model of the impedance of the lesion to be tested can be compared to a pre-established value to determine the result of the comparison of the effective capacitance. For example, if the cells of a tissue are in an inflammatory state, the junction between the cells is loose. The low frequency resistance, ie, for example, the resistance 32 of the model 24, is low compared to healthy cells. Thus, a comparison of the value of this resistance with a pre-established value for healthy non-inflammatory cells may allow to determine the inflammatory state of these cells.

It should also be noted that other parameters of the model may be considered to distinguish the tissue / material of the lesion. However, with these other parameters, it may also be possible to determine an additional item of information about the lesion to be examined. Thus, for example, _{R 2} or the sum _R 1 + _{R 2} of the resistor 26, 32 of the model 24, if the lesion contains a tissue, it may be considered to determine the thickness of the cell structure. To do this, the value R _2, and possibly R _1, is determined, in particular, to optimize the correlation of the measured impedance spectrum with the model 24 simultaneously with the determination of the impedance of the constant phase element. The value R ₂ , or the sum R ₁ + R ₂ , may then be compared, for example in vitro, with the corresponding value predetermined under known conditions. These predetermined values may in particular be stored in a data store.

  As described above, the method can be easily performed in connection with a device that can be inserted into an animal subject, for example, into the vasculature of a human subject.

  By way of example, FIG. 11 illustrates an example system 100 for implementing the method described above.

  The system 100 includes a means 102 for measuring the impedance of a lesion 104, here, for example, blood, a monolayer of confluent cells immersed in a medium 105, and impedance for performing and measuring the method. An electronic control 106 connected to the measuring means 102 for distinguishing the tissue of the lesion 104 as a function of

  Here, the measuring means 102 includes an alternating current generator 108 connected to two electrodes 110, 112 in contact with the lesion 104. The measuring means 102 also includes a device 114 for determining the intensity through the lesion 104 connected to the lesion 104 by two electrodes 116, 118 in contact with the lesion 104. The electronic control unit 106 is connected to the generator 108 and the intensity measuring device 114 so that the impedance of the lesion 104 can be determined, for example, from the measurement of the voltage and intensity at the terminals of the electrodes 110, 112, 116, 118. ing.

  The electrodes 110, 112, 116, 118 are made of, for example, a conductive material such as gold.

  Here, preferably, the measuring means 102 further comprises a medical device 120, here an invasive probe 120, which can be inserted into an animal subject. In this case, the electrodes 110, 112, 116, 118, the AC voltage generator, and the intensity measuring device may be fixed to the medical device. This medical device is described, for example, in French Patent Publication No. 3026631 filed Oct. 3, 2014, the entire content of which is incorporated herein by reference, in particular the description of implantable medical devices including measuring devices. (MEDICAL DEVICE PROVIDED WITH SENSORS HAVING VARIABLE IMPEDANCE).

  In this case, the alternator 108 is configured to conduct current under the action of an electromagnetic field generated by an interrogation unit outside the stent 120, and the armature, eg, the body of a medical device, and It may include an antenna that is electrically insulated from the body of the medical device. Thus, the electrodes may form sensors with various impedances, which vary as a function of the cellular structure covering them. Finally, the electronic control may receive an item of information relating to the impedance between the electrodes, particularly when a magnetic field is generated by an antenna fixed to the body of the implantable medical device 120.

  Thus, the stent 120 may allow the accurate progression of endothelial healing to be confirmed after the stent 120 has been adapted. Indeed, the stent 120, in cooperation with the electronic control, allows the cell structure formed on the surface of the endothelium to essentially replace healthy endothelial cells, inflammatory endothelial cells, smooth muscle cells, and / or platelets. Whether or not to do so can be determined by implementing the method of FIG.

  The invention is not limited to the examples described herein above, but many variations are possible within the scope of the definition provided by the appended claims.

  Thus, for example, at step 22, it is possible to select one model of the impedance of the lesion. In this case, there is no need to perform optimization for many models. Thus, the method is simpler in this case and faster to implement. In particular, if a model is deemed more relevant, it can proceed in this way.

  Furthermore, in some examples described, the tissue / substance distinction is essentially based on a comparison of the calculated effective capacitance with a previously established value of the capacitance. However, as a variant, it is possible to distinguish between the parameters of the selected model of the impedance of the lesion and the tissue / material. However, accurate comparison of the value of the effective capacitance is simple and allows for a reliable tissue / material distinction.

  FIG. 19 illustrates an example of a system 300 created in accordance with aspects of the present disclosure. The system includes a measurement module 301, which may be part of an implanted device, eg, a stent, or may be part of a device for in vitro culturing of cells.

  The measurement module includes at least two electrodes and may be as described above with reference to FIG.

  The system 300 also includes, for example, an internal processor 302 configured to create an impedance spectrum from data from the measurement module.

  System 300 wirelessly transmits data (data from measurement module 301 and / or impedance spectrum determined by internal processing unit 302) to receiver 304, which may be outside the body if the measurements are taken in vivo. An emitter 303 to perform the operation. This transmission may take place under any wireless protocol, for example, especially wireless or infrared, RFID, NFC, Bluetooth®, WiFi. In some embodiments, the transmission may include transmission over one or more wired and / or wireless local networks and / or wide area networks, including the Internet.

The system 300 may include an external processor 305 for computerizing the impedance spectrum (if data from the measurement module 301 is received from the emitter 303) and / or various parameters and effective capacitance C _eff based on the received data. And display means 306, such as an LCD screen, for displaying information related to the cell type and / or condition determined based on the comparison of the reference data and the value representing C _eff . To determine various parameters and effective capacitance, the external processing unit 305 may be configured with information about one or more equivalent circuit models for impedance, and may include at least one parameter of the model, such as the method described above. Can decide. Further, the external processing unit 305 may be configured to select one of the models after determining the parameters of the model as a model for determining the effective capacitance as described above. The external processing unit may perform the selection based on the degree of matching between the equivalent circuit model and the impedance spectrum. The system provides information indicative of the evolution of the healing process based on at least one type and / or condition of cells identified thereby, for example, certain techniques (including positioning of implants such as stents), and / or Or providing information about the current state of the area, such as a healing response or a scarring response, that provides information about changes in the state of the area over time that may reflect a response to the procedure in the area (eg, tissue) obtain.

  The external processor may be a special purpose device including dedicated hardware, such as an ASIC, an EEPROM, or other components, specifically configured to perform the operations of the external processor described above. In another embodiment, the external processing unit is a general-purpose device such as a laptop or desktop personal computer, a server, a smartphone / mobile phone, a personal digital assistant, a tablet computer, or another computer device including a mobile computer device. possible. If the external processing unit is implemented with a general-purpose device, the general-purpose device may include one or more processors and a non-transitory computer-readable storage medium having encoded instructions for execution by the processors (eg, instructions). (Removable media such as registers, on-chip caches, memories, hard drives, optical media, etc.), and the instructions cause the processor to perform the above-described operations performed by the external processing unit. The internal processor may, in some embodiments, be any suitable IC chip or other hardware component with processing capabilities. The external processing unit and the internal processing unit may be located close to each other (eg, in the same room or 5 feet) when the external processing unit is implemented on a server and transmits data via one or more networks or the Internet. Within, or apart from (eg, different parts of a building or complex) or located geographically far from each other (eg, a few miles). You may.

  As a variant, as shown in FIG. 20, part of the processing is performed at a remote server 310 that transmits data, for example, over the Internet.

Example FIG. 25A shows a set of amplitude and phase impedance spectra measured on a cell structure including three cell types, namely, platelets, smooth muscle cells, and endothelial cells, respectively.

Comparative Example First, we use an equivalent circuit model without CPE, consisting of a solution resistance in series with R0Cmix (R0 resistance in parallel with Cmix capacitance) and a double-layer capacitance Cdl in series.

  Next, the Cmix parameters describing the effect of the cell layer on the complex impedance are computed.

  The results of the distribution of Cmix for the two cell types are shown in FIG. 26A. It is possible to distinguish between the two cell types. When a third cell type is added, the three cell types cannot be further distinguished, as shown in FIG. 26B.

  Using a more sophisticated approach and incorporating the CPE elements into the equivalent circuit model, for example using the model 34 shown in FIG. 8A, the six parameters describing the system are: R0, Rinf, Q0, β , Qdl, and α are present.

  These parameters can be computed such that the impedance of the equivalent circuit model best fits the experimental impedance spectrum curve in FIG. 25A.

  This parameter distribution for the three cell types can then be displayed for each parameter, as shown in FIGS. 27A to 27F.

  For each parameter, the three cell types cannot be clearly distinguished, indicating that a linear combination of these parameters cannot provide the distinction of the cell being sought.

Example According to the Present Invention FIG. 28 shows the distribution of values representing the effective capacitances C _eff for three cell types, determined based on equation [8] above.

  It can be seen that all of the three cell types can be clearly distinguished. This accuracy is over 90%. The distinction between cells is significantly improved as compared to FIGS. 27A to 27F.

If the equivalent circuit is one 34 'in FIG. 8B, the _Ceff distribution in FIG. 29 is obtained.

If R0-Rinf is considered large compared to Rinf, equation [8] can be simplified as C _eff = (1−α) / α.

The resulting distribution of C _eff is shown in FIG. It can be seen that the three cell types can still be distinguished with about 85% accuracy.

  The distributions shown in FIGS. 28-30 can serve as reference data for cell type determination.

  For example, the impedance spectrum may be measured under the same conditions as the impedance spectrum of FIG. 25A, and the values of parameters R0, Rinf, Q0, β, Qdl, and α are determined based on this spectrum. This determination may be based on a least-squares fit of the equivalent circuit model 34 of FIG. 8 with the amplitude and phase impedance curves.

Next, if the parameter values R0, Rinf, Q0, and α are known, the effective capacitance C _eff is computerized and this value is compared with the distribution of FIG. 28 to see what cell types correspond Can be determined. For example, a low value of C _eff (nF / cm 2) indicates that the cell is of the first type; a value of about 50 to about 100 indicates that the cell is of type 3, and a value greater than about 100 Indicates that the cells are of type 2.

Examples of Anatomy Techniques As described above, the effective capacitance of an anatomy, including cells, tissues, and / or lesions (including lesions containing cells and / or other materials) is determined based on captured impedance measurements. And may be used to identify the composition of the anatomy (eg, cells and / or tissue of the lesion). However, we have found that using effective capacitance to identify the composition of a lesion or otherwise characterize a lesion may not be the most effective option under all conditions Understand.

  For example, using effective capacitance to differentiate between different cell types in a lesion can be very effective (eg, achieving 95% accuracy), where the measurement of each cell type is Captured under controlled conditions (eg, same temperature conditions, same flow conditions, etc.). The determined effective capacitance distribution for platelets, smooth muscle cells, and endothelial cells under controlled conditions is illustrated by the histogram in FIG. 21A. As shown, the effective capacitance of platelets, smooth muscle cells, and endothelial cells has little overlap. In particular, the effective capacitance of platelets is typically less than about 40 nanofarads per square centimeter, the effective capacitance of smooth muscle cells is typically 40-90 nanofarads per square centimeter, and the effective capacitance of endothelial cells is typically 90 per square centimeter. More than nanofarads. With little overlap, it is possible to use the effective capacitance to reliably distinguish between these different anatomy.

  However, the effective capacitance may be less reliable when discriminating between different anatomy under less controlled or uncontrolled conditions. This may include different temperature conditions, different flow conditions, or other changes. Such variations may be present during measurements in vivo. The determined effective capacitance distribution for platelets, smooth muscle cells, and endothelial cells under uncontrolled conditions is illustrated in the histogram of FIG. 21B. As shown, the effective capacitance of platelets substantially overlaps the effective capacitance of smooth muscle cells. Further, the effective capacitance of smooth muscle cells substantially overlaps the effective capacitance of endothelial cells. The overlap between the effective capacitances of different cell types reduces the performance of cell differentiation techniques that use the effective capacitance of the cell.

  The present inventors have developed techniques for reliably identifying anatomical features (eg, types and / or conditions) such as tissue type and / or composition and / or cell type and / or composition. developed. Such techniques may utilize machine learning. For example, machine learning can be used to interpret and classify EIS measurements to identify the composition of anatomy, such as animal tissues, cell populations, lesions, or other structures or aggregates of biological material. The use of machine learning techniques to identify anatomical features offers many advantages over conventional approaches. The machine learning techniques disclosed herein may provide more accurate results than would be possible through the use of determined effective capacitance under some uncontrolled conditions. For example, some of the trained models developed using the machine learning techniques described herein may identify the composition of some anatomy with 99% accuracy. Furthermore, using a trained model to identify anatomical features may be less computationally intensive than other analysis techniques. For example, using a trained model to identify the composition of the anatomy involves summing a series of operations and operations, including weight values or other values created during training of the model. May only need the equipment to perform. In contrast, identifying the composition of the anatomy based on the effective capacitance of the lesion involves performing a computer-loading process to derive the effective capacitance from the impedance measurement, including adapting the model to the impedance measurement. Equipment may be required.

  In some embodiments, machine learning techniques can be used to identify the relative abundance or concentration of different types of cells or tissues present in a lesion (eg, a blood clot). In this way, it is possible to distinguish between certain types of cells or substances of the same type but with different relative amounts or concentrations, for example blood clots of red blood cells of different relative amounts or concentrations. Thus, in some embodiments, a model may be trained to identify the amount of a particular biological substance (a particular type of cell, such as a red blood cell) in a anatomy. The amount of the substance identified can be an absolute quantity, such as a particular volume or mass of the substance, or other value representing the amount of the substance. In other embodiments, the amount of the identified substance is a relative amount, such as an amount compared to the amount of one or more other substances, including the amount compared to the total amount of the other substance in the anatomy. possible. For example, a ratio that identifies the amount of biological material (eg, red blood cells) in the lesion relative to the entire lesion can be determined. The ratio can be a ratio by volume, a ratio by mass, or any other suitable value that reflects the amount of certain biological material in the lesion.

  In some embodiments, the machine learning techniques described herein are used to quantify the relative amounts or concentrations of different types of cells or tissues that make up or are part of a lesion or other tissue. Can be used. For example, in one embodiment, it may be determined using machine learning techniques that the clot contains 50% red blood cells, 30% fibrin, and 20% platelets. As another example, one embodiment relates to only one type of material, for example, where 50% of the lesion is made up of red blood cells, but does not specifically identify the other 50% of the lesion. One can determine the relative amounts of the substances.

  In addition, the present inventors have realized that direct application of machine learning techniques to tissue and / or cell classification may have inadequate results, and certain methods of utilizing machine learning techniques have been identified. Understand the value of For example, training a model directly using machine learning techniques with only EIS measurements can result in inaccurately low accuracy (eg, less than 80% accuracy) in some environments, for organizations or other models. May yield a trained model that classifies the anatomy of We can see that this reluctantly low accuracy can occur when only raw EIS measurements are used to train the model, and the characteristics received by normal machine learning techniques are irrelevant (eg, correlated). ), But understand that the amplitude and phase points of the correlated EIS measurements can be the cause.

  The present inventors have realized that using derived properties in addition to raw EIS measurements to train a model using machine learning techniques can improve the performance of the resulting model. I have. Derived properties include values generated from analysis of raw EIS measurements, including computer processing performed on raw EIS measurements. Such derived characteristics may include, for example, values representing changes in EIS measurements between frequencies, such as between adjacent frequencies of a set of recovered frequency points. Alternatively or additionally, the derived properties may result from the performance of one or more statistical computations on the raw EIS measurement. Exemplary derived properties that may be used include the phase maximum frequency of the EIS measurement, the n-th quantile of the EIS measurement, the first derivative of the EIS measurement, and the second derivative of the EIS measurement. .

  In addition to the properties present in the EIS measurement (eg, data values that are included in the set of EIS measurements and can be used as properties, as opposed to values resulting from the EIS measurement), the set of properties including derived properties is Training a model using may, for example, result in a trained model that can identify a particular type of tissue in a lesion with up to 99% accuracy.

  The inventors have also recognized that the use of machine learning with EIS measurement data can be hampered by resource constraints on some medical devices that can provide EIS measurements. For example, in some embodiments, an implantable medical device and / or an insertable medical device (including devices described elsewhere herein) have a limited number of devices due to various design constraints. With an EIS sample (eg, 10 samples), it may be necessary to identify one or more features of the anatomy. Such medical devices remain in animal tubing only for a limited amount of time without causing damage and limit the number of samples that can be collected by limiting the time that measurements can be made by invasive probes ( For example, it may include an invasive probe 210). Additionally or alternatively, such medical devices have limited processing power to handle only a limited number of EIS measurements, or have limited bandwidth for communication measurements, or have storage for maintaining the measurements. May be limited or subject to other resource constraints. These constraints on the amount of data that can be collected and provided to the trained model or used to train the model can undermine the effectiveness of machine learning techniques.

  Recognizing this difficulty, we trained a machine learning model that could accurately identify anatomical features when there was only a limited number of samples (eg, 10 EIS samples). Understand the value of the process. An example of such a model training process is illustrated by process 2200 in FIG. Model training process 2200 may include multiple sets of EIS measurements (e.g., 5, 10, 20, 50, 100, 500, 1000 measurements, or more than 1000 measurements per set) each including many samples (e.g., Training data, including 5 sets, 10 sets, 15 sets, 20 sets, 50 sets, 100 sets, 500 sets, 1000 sets, or more than 1000 sets). Here, each sample is a sample related to the impedance at a specific frequency of the applied electric signal. Each set of EIS measurements may characterize a particular biological sample under a particular set of conditions, and in some embodiments, different training sets may correspond to different anatomy. The training process 2200 may identify particular frequencies in multiple sets of EIS measurements that provide the best indication of particular features of the identified biological sample and / or provide data that best distinguishes different anatomy. And Thus, training process 2200 selects a subset of training data corresponding to a particular set of frequencies (eg, a set of ten frequencies), builds a model using the subset of training data, and receives training. It can be used to analyze the performance of the model and determine if the performance is sufficient. If the trained model performs poorly, a new subset of training data corresponding to a different set of different frequencies may be selected and the process may be repeated. After the appropriate combination of frequencies has been identified and a corresponding trained model has been created, the medical device captures the EIS measurements associated with the frequency combination (eg, a set of 10 frequencies) and performs training. Applying the captured EIS measurement to the received model or interpreting the EIS measurement using coefficients (eg, weight values) or rules arising from the trained model that can be used to distinguish anatomy Can identify one or more features.

  As described above, FIG. 22 illustrates an example training process that may train a model to identify one or more features of a tissue and / or cell based on a small number of samples (eg, 10 EIS samples). 2200 is shown. This modeling process 2200 may be performed by the medical device and / or provide the resulting trained model and / or coefficients or rules arising from the trained model to the medical device. It may be performed by a computer system communicating with the medical device. As shown in FIG. 22, the training process 2200 includes a block 2200 for receiving training data, a block 2204 for selecting a subset of training data, a block 2206 for identifying characteristics of the subset of training data, and identification for training. It includes a block 2208 for preparing characteristics, a block 2210 for training the model using machine learning, and a block 2212 for determining whether a performance goal has been achieved.

  At block 2202, the system may receive the training data. The particular composition of the training data may vary depending on the desired features identified. For example, models can be trained to distinguish between different types of tissues and / or cells in a lesion. In this example, the training data may include multiple sets (eg, 20 sets) of EIS samples (eg, 100 samples per set). Each set of EIS samples may be associated with a particular type of tissue and / or cell identified (eg, platelets, smooth muscle cells, and endothelial cells) under certain conditions. The EIS samples in each set may represent the magnitude and / or phase of the impedance of the anatomy at a particular applied frequency.

  At block 2204, the system may select a subset of the training data to use to train the model during one training iteration. The subset of training data may include multiple sets of EIS measurements, and more specifically, each set may include a subset of EIS measurements compared to the originally entered set. These subsets may include, for example, EIS measurements for only certain frequencies of the applied signal.

Thereafter, the identified subset of training data may be used to train the model in blocks 2206-2210. If the performance of the resulting trained model does not match the appropriate performance goal, the system may return to block 2204 and determine a new training data subset different from the previously determined training data subset. For example, a first subset of the training data may include EIS measurements from each of a set of samples corresponding to frequencies f ₁ , f ₂ , and f ₃ . In this example, the system may then determine that the model trained using this subset of training data performed poorly. Thus, the system may return to block 2204 and select a second subset of training data that may include EIS measurements from each of the set of samples corresponding to frequencies f ₁ , f ₃ , and f ₅ .

  It should be understood that, at least in some situations, the measurement data may be at least partially distorted by the presence of noise. Thus, biological material can be sampled multiple times using sensors to statistically reduce the effects of noise on the data. For example, the biological material can be sampled such that at least three spectra are obtained in less than three seconds, at least five spectra are obtained in less than three seconds, or at least ten spectra are obtained in less than three seconds. . By using multiple spectra, the degree of certainty of the model can be increased.

  It should also be understood that the system may use any of a variety of techniques for identifying a subset of the training data. In some embodiments, the system may randomly select a subset of the training data. For example, the system may randomly determine a set of frequencies and select a measurement from training data corresponding to the random set of frequencies. In other embodiments, the system may use a genetic algorithm to intelligently select a subset of the training data. The genetic algorithm may consider the performance of the trained model, for example, using a subset of the previous training data in previous iterations, and if the system determines at block 2212 that the performance goal was not met May be applied.

  At block 2206, the system may identify certain characteristics in the subset of training data for use in training the model. The system may, for example, identify EIS measurements in a subset of the training data as characteristics. The system may also determine one or more derived characteristics that result from the EIS measurement. For example, the system may determine first and / or second derivatives of EIS measurements on a subset of the training data. The first and second derivatives may be calculated in subsets for each set of measurements, and may be calculated in a particular set as derivatives based on EIS measurements. For example, if the measurement set includes 10 samples, the system will determine the first order for each pair of values as a change in amplitude or phase (or both as separate values) between adjacent frequencies in the measurement set. Derivatives can be calculated. In this case, the second derivative may be calculated as the change (in amplitude or phase, depending on the nature of the first derivative) between adjacent values of the set of first derivatives. As another example, the system may determine the phase maximum frequency and / or n quantile of the EIS measurement in a subset of the training data. The n quantile may be a value that divides the area under the curve into n (or nearly n) equal subsets. For example, a curve defining the magnitude of impedance over a range of frequencies may be divided into n (or nearly n) equal sections, and a particular frequency indicating the division between sections is used as a derived characteristic. May be.

  At block 2208, the system may prepare the identified characteristics for use in training the model. Preparing the data may include various functions such as removing noise, removing redundant information, and / or formatting the data. For example, the system may normalize the identified property and / or identify principal components of the identified property using principal component analysis (PCA).

  At block 2210, the system may use at least one machine learning technique to train the model with the identified characteristics. For example, any of a variety of machine learning techniques may be used, including support vector machine (SVM) technology, artificial neural network (ANN) technology, k-nearest neighbor (kNN) technology, or decision tree learning technology.

  At block 2212, the system may determine whether the trained model created at block 2210 matches one or more performance goals. For example, the system may test the machine learning model using training data and / or test data to evaluate the performance of the machine learning model. Assessing performance may include comparing the output created for a given set of inputs with respect to these inputs, such as, for example, comparing a known type of lesion to a lesion diagnosis based on a set of EIS measurements for the lesion. This may include comparing to the predicted output. Such a comparison may be performed one or more iterations of the training, and may generate a value representing the accuracy of the trained model based on the iteratively created model. Performance goals may include, for example, a minimum of accuracy values. If the trained model does not match or exceed the performance goal, the system may return to block 2204 and select a new training data subset. As described above, in some embodiments, if the performance goals do not match, upon returning to block 2204 to determine a particular measurement to select and create a new subset, the system is created at block 2204. Genetic algorithms may be used to learn the choice of frequency over time. For example, the system may start with a plurality of possible combinations of a particular measurement and evaluate the performance of each combination of the particular measurements. In this example, the worst combination of certain measurements may be excluded from consideration, and the best ranked combination of certain measurements is combined to form a new specific measurement combination May be. This process of testing the performance of a particular measurement combination excludes the worst combination of the particular measurements and summarizes the particular combination performed at the top to identify the appropriate particular measurement. Can be repeated until However, in other embodiments, a random selection process may be used at block 2204.

  If, on the other hand, the performance goal is met or exceeded, the system determines that the model has been successfully trained and process 2200 ends.

  After process 2200, a particular set of frequencies to use for EIS measurements that provide appropriate discrimination between anatomical structures was identified, and the model was trained to use these frequencies to distinguish anatomical structures. . This frequency and model may be saved after process 2200. Further, the medical device (including implantable and / or insertable devices described elsewhere herein) may be configured to measure the impedance of the anatomy at these frequencies; The system (the medical device itself, or a medical device in combination with another computing device such as that described above in connection with FIG. 2) may be used to analyze the EIS measurement to perform identification or characterization. Configured to identify the anatomy, identify the composition of the anatomy, or otherwise characterize the anatomy using the trained model and / or coefficients or rules arising from the trained model. obtain.

  One skilled in the art will recognize that in some techniques for training a model, the model learns various weights or other values (also called coefficients) for use in computer processing that produces a particular output from the input. You will understand that In such a case, these sets of weight values can be said to represent the model, along with information about the computer processing performed on the inputs using the weight values. Thus, in some embodiments, as a result of the process of FIG. 22 (or other processes described herein that result in a trained model), to configure the device for subsequent use by the device, A set of available weight values is created. For example, the model trained as a result of process 2200 of FIG. 22 may identify and / or classify a lesion (or other anatomy of interest) based on input characteristics determined from EIS measurements on the lesion. It can result in a model that makes it possible. The model may be represented as a set of weight values that allow one or more features of the lesion to be identified from the characteristics. By configuring the device to perform computer processing using these weight values, the device generates a numerical value from, for example, EIS measurements and weight values that match a particular feature or set of features. By making the determination, a feature can be created based on the input characteristics. Performing such computer processing may be less computationally intensive than other forms of analysis that may be performed on the characteristics entered to create the feature, and thus take less time or processing than alternatives Resource requirements may be reduced.

  It should be understood that various alternatives to process 2200 may be made without departing from the scope of the present disclosure. In process 2200 described above, at block 2206, the same set of characteristics was selected at each iteration based on the selected subset of training data. In some cases, this set of properties may be inappropriate to reach performance goals regardless of which training data subset is selected, or a set of properties that may provide a more accurate analysis May exist. In some embodiments, the process 2200 optimizes or improves certain portions of the training data used to train the model, as well as optimizes certain characteristics used to train the model. Or may be improved. For example, after the system determines that the trained model does not match the performance goal in act 2212, or the system has reached a performance plateau where the trained model does not meet or exceed the performance goal. If it is determined (or a performance plateau has been reached even if the performance meets or exceeds the goal), the system proceeds to block 2210 in addition to (or instead of) changing a subset of the training data. The particular characteristics used to train the training may vary. The selection of the property may be driven by a similar genetic algorithm, as described above in connection with the selection of the value of the measurement. Thus, the system may iterate through multiple, different combinations of properties until a combination of properties that provides the desired performance is identified.

  Example results obtained from using process 2200 to create a trained model that discriminate between various cell types based on 10 EIS measurements are shown in FIG. 23 and FIG. Show. These confusion matrices are tables that allow direct visualization of the results of the model applied to a particular class of sample. Each row of the matrix represents an example of a predicted class, and each column represents an example of an actual class. The diagonal values represent the case where the model accurately predicts the labeled class. The diagonal values represent the probability that the corresponding labeled class will be correctly predicted. Off-diagonal values represent cases where the models confuse classes with each other, with corresponding probabilities. The model distinguishes between the following five classes: 1-bovine aortic endothelial cells (BAEC), 2-bovine aortic smooth muscle cells (BAOSMC), 3-platelets, 4-empty, 5-intermediates. Trained for. FIG. 23 shows the results from applying the training data to the trained model, and FIG. 24 shows the results of applying the test data to the trained data.

  In some embodiments, the cell differentiation method 10 of FIG. 4 may be configured at a particular frequency at which EIS measurements are made, rather than making continuous or pseudo-continuous measurements through the spectrum. This particular frequency may be the frequency that, when analyzed, creates an impedance value that provides the clearest distinction between different biological materials. In some embodiments, the clearest distinction may be an impedance spectrum that has little overlap between the spectra, and in other embodiments, the clearest distinction may be a value that can be used to identify a biological material. Or it may be the least overlap or similarity in a range of values. For example, in some embodiments used to determine effective capacitance and to identify or determine one or more characteristics of a biological material, the most distinctive distinction relates to different types of biological materials. There may be minimal overlap or similarity in the value or range of values for the effective capacitance.

  FIG. 24A illustrates another confusion matrix in which the true markers are shown as rows and the predicted markers are shown as columns. In this case, four classes were considered: "empty", "mix", "red", and "white". Each class represents a different type of clot. For example, the class “red” represents a type of clot rich in red blood cells, the class “white” represents a type of clot rich in fibrin, and the class “mix” represents a type of clot rich in fibrin and red blood cells. It represents a type of mochi, and the class “empty” represents the case where no clot is present. When represented by a matrix created based on 3000 EIS measurements, a clot of class "red" is predicted by a model trained with 100% probability and a class "white" has a 94.5% probability And so on. In contrast, a clot labeled "white" is incorrectly predicted as "mix" with a 10.5% probability and a clot labeled "mix" with a 5.5% probability. It was incorrectly predicted as “white”. The data set for the EIS measurement used to create the matrix of FIG. 24A is shown in FIG. 25B. The top chart illustrates the data points representing the amplitude of the impedance versus frequency for the different classes. The lower chart illustrates data points representing the phase versus frequency of impedance for different classes.

  FIG. 22 relates to techniques for training a model to determine anatomical features based on input impedance measurements and / or to determine frequency / characteristics as part of learning to determine these features. Although described above, it should be understood that embodiments are not limited to the use of the process of FIG. 22 to learn the relationship between impedance measurements and one or more features of the anatomy. For example, as discussed further below in connection with FIG. 15C, in some embodiments, the impedance measurement for the anatomy and the recommended treatment for the anatomy (eg, for a lesion when the anatomy is a lesion). Model can be trained to learn the relationship between In some such embodiments, as discussed in connection with FIG. 15C, the system provides a therapy recommendation for a anatomy detected by the device without performing intermediate steps to identify or characterize the anatomy. To determine, the impedance measurements collected by the devices described herein may be used. In such embodiments, training the model to learn the relationship between the impedance measurement and the treatment recommendation may include identifying frequency and / or characteristics, as discussed by FIG. . Further, as discussed elsewhere herein, such embodiments for generating a treatment recommendation based on impedance measurements may be used, among other types of recommendations, for a treatment option to use from a set of treatment options. Recommendations (eg, different tools to use), recommendations on how to make this treatment option (eg, how to activate the tools created before and / or during the execution of the treatment), or recommendations not to treat A treatment recommendation can be made.

Methods of Operating Medical Devices Examples of medical devices, sensors, and methods of detecting lesional tissue / substance are described in more detail above with respect to FIGS. Examples of techniques that can be performed by such medical devices and / or that can be performed by operating the medical devices are described below in connection with FIGS.

  FIG. 12 illustrates a process 1200 that may be performed, for example, by a medical device operating in accordance with some techniques described herein. The example medical device of FIG. 12 may be a medical device where the invasive probe may include only one sensor, which may include one or two electrodes. As is clear from the above discussion, a limited amount of information about the lesion can be determined from one sensor when compared to multiple sensors placed along the invasive probe (in the example of FIG. 3). . In the example of FIG. 12, the sensor of the invasive probe may be a therapeutic device, such as a suction catheter and a stent-type thrombus collector, and / or a treatment device such as a guidewire inserted prior to insertion of the suction catheter or stent-type thrombus collector. The device may be located. The medical device may generate a treatment recommendation based on the characteristics of the lesion determined using the sensor.

  The process 1200 begins at block 1202, where a sensor attached to a guidewire operates to detect one or more features of a lesion proximal to the sensor. Prior to the start of process 1200, an invasive probe of a guidewire, of which the sensor is a part, may be inserted into the vasculature of the animal and moved proximal to the predicted location of the lesion. The sensor is then activated to detect when the sensor contacts the lesion. Lesion contact can be determined by evaluating the change over time of the value output by the sensor. For example, the sensor may output one value upon contact with blood, which may be the case when the sensor is placed in the center of a blood vessel in an area not blocked by a lesion. If the invasive probe moves before contacting the lesion, the value output by the sensor may change after the contact is made. In this way, the location of the lesion may be determined using one sensor. Further, in some cases, the sensor may determine the length of the lesion, for example, by continuing to advance the invasive probe until the sensor no longer contacts the lesion and the output value returns to a value associated with the blood contacted. Can be operated.

  In the example of FIG. 12, using only one sensor, the medical device may be unaware of the composition of the lesion and may provide a treatment recommendation as to which treatment option may be best for treating a particular lesion. In some cases, it cannot be created. However, the medical device may be able to provide information about the progress or success of the treatment and may be used to determine whether the selected treatment option has been successfully performed. Based on this information, the medical device may formulate treatment recommendations as to whether the treatment being given should be changed to another treatment.

  In one treatment protocol that may be implemented in embodiments such as FIG. 12, a suction catheter may be used as a first option for treating a lesion. Thus, at block 1204, an aspiration catheter is inserted into the vasculature until it is proximal of the invasive probe of the guidewire and thus proximal of the lesion. In some embodiments, the guidewire may not be inserted first, but rather the suction catheter may be inserted at block 1202 until it is positioned proximal to the lesion. In such a case, the sensor may be a component of the suction catheter. Embodiments are not limited to this mode.

  At block 1204, after placing the suction catheter proximal to the lesion, the suction catheter is activated to attempt to aspirate the lesion into the catheter. After a period of time, the sensors on the guidewire and / or suction catheter may be activated to determine whether the suction catheter has an effect on the lesion. Some lesions, such as hard lesions, may not be able to be aspirated using a suction catheter. Other interventions (such as stent thrombus collectors) may be used for these lesions. Thus, at block 1204, in addition to actuating the aspiration catheter for an attempt to aspirate, the sensor may be actuated to determine whether the lesion has changed. This means, for example, that the sensor is placed in the lesion before the start of aspiration, for example in the part of the lesion closest to the suction catheter, and after a certain time the value output by the sensor is This can be done by determining whether it represents not contacting (rather, for example, in contact with blood).

  During actuation of the suction catheter (and possibly as a result of the actuation), if the sensor is no longer in contact with the lesion, a determination that the lesion is being aspirated may be made at block 1206. In this case, a treatment recommendation may be generated and output indicating that the suction catheter appears to be successfully treating the lesion and that continuous operation of the suction catheter is recommended. In the example of FIG. 12, the process 1200 ends thereafter. However, in some embodiments, a continuous determination as to whether the aspiration catheter continues to successfully treat the lesion may be made over time, or when the lesion is It should be understood that a determination of whether or not the aspiration was complete can be made.

  However, if the value output by the sensor does not change during the aspiration, indicating that the aspiration has no effect on the lesion, the aspiration catheter is no longer recommended and instead, A treatment recommendation in which options are recommended can be created and output. In the example of FIG. 12, the second option for treating a lesion may be a stent-type thrombus collector. Thus, at block 1208, a recommendation to use a stent clot collector may be output. At block 1210, the stent clot collector may operate to treat the lesion by removing the lesion with a stent clot collector. For example, a stent-type thrombus collector may be inserted until it is located proximal to the lesion. In some embodiments, as described above, the sensor from which the detection is made may be a component of the guidewire or may be remote from the treatment device. In such a case, the stented thrombus collector may be inserted along the guidewire after removal of the aspiration catheter until the stented thrombus collector is positioned proximal to the lesion (or the removal of the guidewire). Later, it may be inserted along a macrocatheter inserted along a guidewire). As another example, the sensor may be integral with the stented thrombus collector and may detect when the stented thrombus collector is positioned proximal to the lesion. The medical device may, through the values provided using the sensors, make a treatment recommendation regarding the placement of a stent thrombectomy to remove the lesion. For example, sensors can be used to detect when an invasive probe has passed a lesion and when the distal end of the invasive probe has been placed behind the lesion, as described above. It is best to position the stent of the thrombectomy across the lesion so that one end of the stent protrudes beyond the lesion to help ensure that the stent has completely captured the lesion Can be Thus, by activating the sensor to detect the backside of the lesion and recommending that the stent or thrombus collector be inserted until the stent or sensor reaches the lesion, a treatment recommendation for proper placement of the stent can be made.

  After activating the stent clot collector to remove the lesion at block 1210, the process 1200 ends.

  FIG. 13 illustrates an example of a method of operating a medical device to create a treatment recommendation for a lesion, according to another embodiment. In the embodiment of FIG. 13, the invasive probe may include multiple sensors located along the outside of the probe, such as the example of FIG. 3 described above. As will be appreciated from the above, with the arrangement of such sensors, several different characteristics of the lesion, including the composition of the lesion, can be determined. The composition of the lesion may represent a different biological material present in the lesion, such as a different tissue or cell, or other biological material, such as a plaque-like material. In some such embodiments, for example, each sensor (e.g., two electrodes of each sensor) may be in contact with the biological material of the lesion, where some of the sensors may be more active than others. Contact with different pathological biological materials. Each sensor may then be operated according to the techniques described herein to determine the impedance spectrum of the biological material that the sensor has contacted. The set of impedance spectra can then be used to determine the composition of the lesion, for example, by identifying different biological materials present in the lesion. This compositional information may be similar to the information that can be determined from performing a histological examination for the lesion. From the identification of different impedance spectra for the lesion and / or different biological materials present in the lesion (eg, different tissues or plaque-like materials), the characteristics of the lesion as a whole, eg, identify the type of lesion (eg, diagnosis) Can be determined.

  For example, by performing an EIS process on biological materials with different lesions, the following cells or tissues: platelets, fibrin, thrombus, erythrocytes, leukocytes, smooth muscle cells, elastic fibers, outer elastic lamina, internal elastic lamina, loose connective tissue , Endothelial cells, or any of the intima, media, or adventitia of any other tissue can be determined to be present in the lesion. In addition, by performing an EIS process on the lesion, the relative amount of each of the cells or tissues present can be determined. As a simple example, it can be determined that the lesion is composed of 50% red blood cells, 30% fibrin, and 20% platelets. From this information, the lesion may be classified as one particular type of lesion from a set of lesions, for example, by diagnosing the lesion as one of the other types of lesions.

  The process 1300 of FIG. 13 begins at block 1302, where an invasive probe of a medical device is inserted into the vasculature of an animal subject and operates to detect one or more features of the lesion, including the composition of the lesion. I do. Based on features including this composition, the medical device selects a recommended treatment option at block 1304. The medical device may select the treatment options in any suitable manner, including according to the techniques described below in connection with FIGS. 14-15B.

  The treatment option selected may be selected based on the composition of the lesion. For example, if the composition of the lesion indicates that it is composed of smooth muscle tissue rather than thrombus, the medical device may determine that stent implantation is the treatment to be recommended. This may be because the lesion is not made up of cells / material that can be extracted, and instead is a tumor in a blood vessel. As another example, if the composition of the lesion indicates a soft lesion, for example, a soft lesion created from a freshly formed thrombus, the medical device may recommend a suction catheter. This may be because soft lesions can be aspirated. As a further example, if the composition of the lesion indicates a hard lesion, e.g., a hard clot, the medical device may use a stent-type thrombectomy device because the likelihood of successful suction of the hard lesion is low. Can recommend.

  Once a treatment is recommended at block 1304, the medical device may monitor the performance of the treatment option selected at block 1306. The medical device monitors treatment using one or more sensors, for example, one or more sensors whose characteristics have been determined in block 1302, or one or more sensors of a treatment device operative to provide treatment. obtain. For example, in some embodiments, following the recommendation of block 1304, the clinician may insert another device into the vasculature of interest (eg, a suction catheter, stent thrombus collector, etc., as needed), This other device may include an invasive probe having a sensor arrangement as described herein. In such an embodiment, the medical device may use the sensors of the invasive probe of another device to monitor the performance of the treatment.

  Monitoring the therapy at block 1306 may provide information regarding the status and / or progress of the therapy. For example, if the treatment is performed using a suction catheter, the monitoring may provide information about the extent to which the lesion has been aspirated and / or the amount of residual disease that has been aspirated. This progress can be achieved by, for example, contacting the sensor with the rest of the lesion and periodically or possibly expanding a structure (eg, the stent-like mesh of FIG. 3) to determine the extent of the remaining lesion. It can be monitored by the instrument. After the decision is made, the structure may be removed and the aspiration of the lesion may continue. On the other hand, if the treatment is performed using a stent thrombus collector, monitoring can provide information on the extent to which the stent has merged with the lesion during stent expansion. For example, by monitoring the sensors along the outside of the stent (e.g., using the placement of the sensors on the stent, as in the example of FIG. 3), each portion of the stent corresponding to each sensor is located within the lesion. A determination of whether or not fully inflated may be made. This determination may be made in any suitable manner, including by monitoring the change in value provided by each sensor over time and determining when the value for each sensor has stopped changing. If each sensor stops changing value, this means that the interaction between the lesion and the stent will not change further, so that the stent will expand completely within the lesion and the lesion will Can indicate that they are united.

  Making such a determination can aid in the performance of the treatment of the lesion. Thus, at block 1308, information regarding the status of the treatment is output by the medical device via the user interface for presentation to the clinician. Further, at block 1310, the medical device may formulate one or more treatment recommendations regarding how to perform the treatment. For example, if the medical device determines that the lesion has fully coalesced with the stent during actuation of the stented thrombus collector, as described above, the medical device may output a treatment recommendation to initiate stent removal. .

  After a successful treatment, the process 1300 ends.

  An example of monitoring treatment is provided in the context of developing a treatment advisory, but it should be understood that similar techniques may be used to result in error messages or other messages to the clinician regarding the condition of the treatment. It is. For example, if a sensor on a treatment device indicates the presence of a lesion at a time after the sensor no longer detects the lesion, the medical device may indicate that the treatment device is improperly placed or that the lesion has disappeared. You can decide. This may indicate that the device needs to be repositioned, or that the lesion has become embolized, possibly causing more problems. A message to the clinician via the user interface may indicate such a potential problem.

  Further, although the example of FIG. 13 describes a method of activating a medical device to provide treatment recommendations related to the selection of a first treatment and subsequently to a method of performing the treatment, an embodiment is described herein. It should be understood from the above that the present invention is not limited to this. For example, in some embodiments, a medical device may include one or more of the sensors described herein and may provide a method of operating the device without making initial recommendations for using the device. Operable to provide a therapeutic recommendation for For example, a stent-type thrombus collector or aspiration catheter may include one or more sensors for generating data regarding the status or performance of a therapy, as described above, and may generate a therapy recommendation. As another example, a guidewire for the treatment of chronic total occlusion (CTO) may create information about the tissue / substance that the sensor has contacted and may make a treatment recommendation. In the CTO approach, a guidewire can be inserted through the smooth muscle tissue or plaque of a blood vessel if a solidified thrombus cannot be punctured. Based on the detected characteristics of the tissue / material contacted by the sensor, when the guidewire is placed in contact with the smooth muscle tissue and can be advanced, and when the guidewire is advanced through the endothelial tissue and is behind the lesion A treatment recommendation to return to the vessel of another can be made. Further, in some embodiments, one or more measurements of the thickness or other characteristics of the smooth muscle tissue of the vascular wall may be more useful at risk of piercing the tissue than a guidewire is advanced through the tissue. May go. For example, if the measurement represents the thickness of the smooth muscle tissue on one side of the invasive probe of the guidewire, this may indicate that the invasive probe is at risk of piercing the vessel wall. A treatment recommendation to advance the guidewire more slowly and / or retrieve the guidewire may be made, or another recommendation may be made.

  Those of skill in the art will understand from the discussion herein that there are various ways in which a medical device can be configured to generate a treatment recommendation based on the characteristics of the lesion and / or the state of the treatment. . 14-15B illustrate one example of a technique that may be used to create a treatment advisory.

  FIG. 14 illustrates a process 1400 that may be performed by a medical device in some embodiments for creating a treatment advisory.

  Process 1400 begins at block 1402, where a medical device receives one or more features of a lesion. The medical device may include, for example, one or more sensors included in an invasive probe of the medical device and / or another component (eg, a lesion analysis facility) that creates features based on data provided by the sensors. ), The feature may be received from a component of the medical device. This feature may include, in some embodiments, the composition of the lesion. Additionally or alternatively, the characteristics may be determined based on the location of the lesion in the body, one or more dimensions of the lesion (eg, length, thickness, etc.), the temperature of the lesion, or other types of sensors described above. May contain information.

  At block 1404, the medical device compares the feature received at block 1402 with one or more conditions for one or more treatment options. The medical device may be configured with information regarding a plurality of different available treatment options, each of which may be associated with one or more conditions associated with one or more characteristics of the lesion. For example, medical devices may include one or more conditions for treating a lesion by implanting a stent, one or more different conditions for using a suction catheter, and one more for using a stent-type thrombectomy device. It can be configured with the above different conditions. Examples of such conditions related to lesion composition are described above in connection with FIG.

  The medical device may compare the plurality of conditions with the characteristics of the lesion to determine which conditions are met. In some embodiments, the set of conditions for treatment options are mutually exclusive such that one lesion may only match one set of conditions, and thus only one treatment option may be selected. possible. In other embodiments, the set of conditions may not be mutually exclusive, and the medical device may select the option that matches the most corresponding condition, or the option that the corresponding condition most closely matches (e.g., If a condition is associated with a value in a range, which treatment option is recommended, by identifying the condition whose value most closely matches the range, for example, by being in the middle of the range) Can be determined.

  At block 1406, based on the comparison, the medical device may output a treatment option recommendation via the medical device user interface, and the process 1400 ends.

  Although the process 1400 has been described in connection with the creation of an initial treatment recommendation for treatment of a lesion based on the characteristics of the lesion, one of ordinary skill in the art will appreciate that during the performance of the treatment as described above in connection with block 1310 You will understand how to extend this technology to the development of treatment recommendations. For example, in some embodiments, comparing one or more conditions for a particular parameter of treatment, such as the rate at which a stent-type thrombectomy device withdraws a stent, with a lesion characteristic (eg, lesion composition). Based on that, the medical device may output a recommendation for the parameter.

  One skilled in the art will appreciate that there are many ways to set conditions on treatment options that may be used in connection with a process such as process 1400 of FIG. For example, the values for lesion characteristics used as conditions may be determined after at least some experiments to determine the correspondence between these values, lesion type, and treatment success with various treatment options. It can be hard coded into the device. However, we recognize and understand the benefits of a system that learns the relationships and conditions based on information, among other information, on the characteristics of the lesion and the successful treatment of the lesion. For example, a machine learning process, for example, a machine learning process that may include feature extraction and / or classification, may be implemented in some embodiments.

  15A-B are examples of a machine learning process that may be performed in some embodiments. FIG. 15A illustrates a process that may be performed by a medical device, and FIG. 15B illustrates a process that may be performed by a computing device (eg, a server) communicating with a plurality of different medical devices.

  The process 1500 of FIG. 15A begins at block 1502, where the medical device creates information regarding the characteristics of the lesion. In blocks 1504 and 1506, the medical device may make recommendations for treatment options based on a comparison of the conditions for the treatment options and the characteristics of the lesion, monitor the progress of the treatment, and monitor the condition of the condition throughout the treatment. Information may be created. These operations of blocks 1502-1506 may be implemented in a manner similar to that described above in connection with FIGS. 13-14, and thus are not further described for brevity. Further, at block 1506, the medical device may create information regarding the prognosis of the treatment. The prognosis of treatment may indicate whether the treatment of the lesion was successful, whether the lesion was removed and removed in the subject's body, whether multiple treatments were needed, and other information that indicates prognosis. Prognostic information may be created using the sensors of the medical device, as will be apparent from the above. For example, using data generated by the accelerometer on the medical device's handle, the medical device may determine whether it has acted multiple times to remove the lesion. As another example, if the sensor detects a lesion and then stops detecting the lesion, as described above, this indicates that the lesion has moved in the subject, including that the lesion has been removed and has become an embolism. Index.

  At block 1508, the information created at blocks 1502 to 1506 is transmitted from the medical device to the computing device via one or more wired and / or wireless communication connections and / or networks, including the Internet. The computing device may be geographically remote from the medical device in some embodiments. At block 1508, after the transmission of block 1506, the medical device receives one or more updated conditions for the treatment option from the computing device (eg, via the network over which the information was transmitted at block 1508). . The updated condition may identify new values for evaluating the condition with respect to lesion features. The medical device itself may be used to create one or more treatment recommendations through consideration of one or more updated conditions, for example, in the context of a process such as that described above in connection with FIG. It can be set to apply the updated conditions described above. After the medical device has been configured with the updated conditions, the process 1500 ends.

  FIG. 15B illustrates a computer that performs the process of learning a report on the treatment of a lesion to create conditions for use in selecting a treatment recommendation, for example, via a process such as described above in connection with FIG. 2 illustrates a process that the apparatus may perform. Specifically, in the example of FIG. 15B, the computing device analyzes the report on the treatment of the lesion in relation to the information on the characteristics of the lesion, and reports the success (and / or failure) of the treatment and the characteristics of the lesion. Identify the relationship between Through the identification of such a relationship, conclusions may be drawn as to which treatment option is best for a particular type of lesion, based on which conclusions a treatment recommendation may be made as in the example of FIG. Then, a treatment for this lesion can be created based on the characteristics of the particular lesion. Similarly, as described above, based on information regarding the status or performance of the treatment, recommendations regarding how to perform the treatment (eg, time or speed withdrawing the stent during retrieval of the stent) may be determined. Although the example of FIG. 15B is described in the context of creating conditions for the initial selection of treatment options to use for a lesion based on the characteristics of the lesion, those skilled in the art will be able to make recommendations regarding how to perform the treatment. How to extend the techniques used will be understood from the description below.

  The present inventors have recognized that the creation of such conditions and the identification of the relationship between treatment success / failure and lesion characteristics can be preferably determined using a machine learning process. I understand. Various machine learning algorithms are known in the art and may be adapted for use in this situation. Some machine learning algorithms may operate based on feature extraction and classification techniques, where a grouping (classification) for multiple units is identified, an analysis of the unit's properties is performed, and which properties and / or Determine whether the values of the properties of the most closely correspond or predict exact membership in the group. Based on these identified properties, subsequently received unclassified units having that property are grouped / grouped based on a comparison of the property / value for each group with the property and / or property value of the unclassified unit. It can be "classified" into one of them. In some machine learning applications, groups / classifications may be identified manually during setup of the machine learning process. Additionally or otherwise, the group / classification may be, for example, the creation of a new group / classification if the machine learning process notices through this analysis that the new grouping may well characterize some units. Via a machine learning process over time. A full discussion of machine learning is outside the scope of this document and is not necessary for an understanding of the technology herein. Those of skill in the art would understand how to implement machine learning techniques for use with the information and for the purposes described herein.

  Here, a group may be defined as a treatment option or treatment prognosis, the example of FIG. 15B being described in this context. In this case, groups may be defined by lesion characteristics and / or treatment status. In this case, if the characteristics of the lesion and / or the state of the treatment match the characteristics of the group, a corresponding treatment option may be selected for output. Additionally or alternatively, in some embodiments, the groups are associated with different types of lesions, each type having one or more features or ranges of features distinct from other types, and / or the condition of the treatment. And thus these different groups may relate to a particular treatment option or method of operating the treatment device. In this latter case, if a feature associated with a particular lesion or treatment condition matches a certain group, the corresponding treatment recommendation of that group may be selected for output.

  The process 1520 of FIG. 15B begins at block 1522, where a learning facility implementing one or more computing devices receives, over time, a plurality of reports regarding treatment of a lesion with a medical device. The medical device can be a medical device that operates according to the embodiments described above. The report may include information regarding the characteristics of the treated lesion, eg, one or more lesions. The report may also include information about how the lesion was treated, for example, one or more treatment devices that were activated to treat the lesion and how these lesions were treated. Information about the prognosis of the treatment, such as whether the treatment was successful, whether multiple treatments were required, whether the lesion was removed and embolism, or other prognosis, may also be included in the report.

This report may include information determined by one or more sensors of the medical device, including examples of sensors and types of information described above. As described above, various types of sensors may be included in embodiments, including one or more electrical, mechanical, optical, biological, or chemical sensors. Specific examples of such sensors include an inductance sensor, a capacitance sensor, an impedance sensor, an EIS sensor, an electrical impedance tomography (EIT) sensor, a pressure sensor, a flow sensor, a shear stress sensor, a mechanical stress sensor, and a deformation sensor. , temperature sensor, pH sensor, a chemical composition sensor, (e.g., O ₂ ions, biomarkers or other compositions), the acceleration sensor, and motion sensors, and the like. It should be understood that various types of features or other information may be created from these sensors. Any of this information may be included in a report to create a condition associated with a treatment recommendation and may be used in process 1520. For example, as described above, an accelerometer located in the handle of the medical device may track the movement of the medical device and determine whether multiple treatments have been taken to treat the clot. Can be used for As another example, a force sensor may represent the force with which the stented thrombus collector is withdrawn, or the set of impedance sensors may be one or more sensors of the stent of the stented thrombus collector. Based on whether the detected impedance varies over time during the extraction, it may be determined whether the lesion has partially or completely detached from the stent during retrieval. Those of skill in the art, from the above discussion, will understand the different types of data that can be produced by medical device sensors for inclusion in such reports.

  The report may also include information that can be entered by a clinician, or information that can be retrieved from another system that the medical device can use simultaneously. For example, the report may include information regarding the location of the lesion in the subject's anatomy, such as whether the lesion is in the cranial artery, femoral artery, pulmonary vein, common bile duct, or other duct. This information may be entered by the clinician via a user interface or may be retrieved from another system, such as, for example, an angiographic device.

  Optionally, the report may include information about the patient, such as age, medical history, and demographics.

  The report received at block 1522 may be received over time from a plurality of medical devices that may be geographically distributed. By receiving these reports and the contents of these reports, a set of conditions and treatment recommendations over time that define recommendations or best practices can be created.

  Thus, at block 1524, the learning facility may identify the characteristics of the lesion (and / or the method of operating the treatment device), options for treating a lesion having these characteristics, and a relationship between successful treatments. Then, analyze the information in the report. Based on this analysis, the learning facility may learn the relationships between these pieces of information. Such a relationship may indicate when a particular treatment option was a success or failure, or for what type of pathology a different treatment option was a success or failure. In at least some of the embodiments in which information about the patient is obtained, the learning facility may determine the relationship between the characteristics of the lesion, options for treating a lesion having these characteristics, and successful treatment based on the patient's information. Can learn. The model can be trained to learn which specific pieces of information of all the information obtained about the patient may affect the probability of successful treatment. For example, a trained model may identify that a particular treatment may have different probabilities of success depending on the age of the patient, even if all of the features of the lesion are equal. Thus, different treatment recommendations may be provided for two patients with the same lesion but different ages. As another example, a trained model can reduce the number of treatments applied to a subject with a particular condition in the past, even if the type of lesion is the same. One may learn that it is less likely to be successful compared to unaffected subjects.

  Based on this analysis at 1524, the learning facility (via the process of characterizing and classifying the machine learning process) creates, at block 1526, conditions for each of the treatment options. This condition may be related to the characteristics of the lesion so as to represent a different characteristic or range of characteristics for the lesion that can be successfully treated with each treatment option. For example, the condition may be viscoelastic of the lesion such that one range of viscoelasticity may be associated with treatment using a suction catheter and another range of viscoelasticity may be associated with treatment using a stented thrombus collector. It may relate to a range of values for elasticity. In this method, if a lesion with a particular viscoelasticity is detected, using a comparison with these conditions (as in the process of FIG. 14), which treatment option is recommended for the particular lesion Can be determined.

  At block 1528, if conditions are created at block 1526, the conditions may be configured such that the medical device uses these conditions to create a treatment advisory as described above in connection with FIG. 15A. Can be distributed to the medical device. Once the conditions have been distributed, the process 1520 ends.

  Although process 1520 is discussed in FIG. 15B as another process, in some embodiments, receiving reports and determining conditions is repeated over time, including at continuous or discrete intervals. It should be understood that this can be a process. Thus, in some embodiments, the process 1520 may be performed multiple times, or after the distribution of the conditions in block 1528, the learning facility may receive additional reports and continue the learning process. Control may return to block 1522.

  In some alternative embodiments, the creation of a treatment advisory may be performed using machine learning techniques, based directly on measured data obtained from one or more sensors of the type described above. Developing treatment recommendations based directly on the measured data can be done without the need to first characterize or identify the lesion or lesion type. Compared to other approaches that use the measured data to identify the nature of the lesion (such as type or composition) and then identify appropriate treatment recommendations, this approach eliminates intermediate steps that characterize the lesion Enable. This may reduce the time and computational resources required to make a treatment recommendation.

  FIG. 15C illustrates a process that may be performed by a computing device to perform a process of learning a report on treatment of a lesion to create a condition for use in selecting a treatment recommendation. This condition may relate to the EIS measurement and / or properties determined from the EIS measurement (eg, properties present in the measurement and / or derived properties). The process of FIG. 15C may be used to train a model to identify one or more relationships between EIS measurements or characteristics and treatments for the pathology for which the measurement / characteristic is associated. With the relationship between measurement / characteristics and treatment, training a model in this manner may allow the creation of treatment recommendations for the lesion without having to diagnose or identify the lesion.

The process 1540 of FIG. 15C begins at block 1542, where a learning facility running on one or more computing devices over time reports containing measurement data obtained using the techniques described herein. Receive. Examples of measurement data include, but are not limited to, EIS measurements. Each set of data may include any suitable number of EIS samples, each of which may represent impedance information obtained at a particular frequency. Thus, each set of measurements can be interpreted as a spectral response of a particular type of lesion. In certain cases, measurement data obtained via sensors other than impedance sensors may be used in process 1540. Examples of such sensors include inductance sensors, capacitance sensors, electrical impedance tomography (EIT) sensors, pressure sensors, flow sensors, shear stress sensors, mechanical stress sensors, deformation sensors, temperature sensors, pH sensors, chemical compositions. object sensor, (e.g., O ₂ ions, biomarkers or other compositions), the acceleration sensor, and motion sensors, and the like.

  The measurements received in the report may include measurements taken during the stage of diagnosis or treatment when the sensor first contacts the lesion, or when measurements are being collected for the lesion. Such a measurement can be used by the techniques described elsewhere herein to identify or diagnose a lesion. Additionally or alternatively, the measurements may include measurements collected during a treatment, for example, during extraction of a lesion using a stent thrombectomy device or during another procedure.

  The report may also include information about how the lesions were treated, for example, one or more treatment devices that were activated to treat the lesions and how these lesions were treated. Information about the prognosis of the treatment, such as whether the treatment was successful, whether multiple treatments were required, whether the lesion was removed and embolism, or other prognosis, may also be included in the report.

  The report received at block 1542 may be received over time from a plurality of medical devices that may be geographically distributed. By receiving these reports and the contents of these reports, a set of conditions and treatment recommendations over time that define recommendations or best practices can be created.

  Thus, at block 1544, the learning facility analyzes the information in the report to identify a relationship between the measured data, options for treating a lesion exhibiting the features presented in the measured data, and a successful treatment. . Based on this analysis, the learning facility can learn the relationships between these pieces of information. Such a relationship may indicate when a particular treatment option was successful or unsuccessful, or for which type of lesion a different treatment option was successful or unsuccessful.

  Based on this analysis at block 1544, the learning facility creates conditions (via the process of character extraction and classification of the machine learning process) at block 1546 for each of the treatment options. Blocks 1546 and 1548 may operate in substantially the same manner as blocks 1526 and 1528 of FIG. 15B.

  As discussed elsewhere herein, one embodiment of generating a treatment recommendation based on impedance measurements in accordance with FIG. 15C is based on the type of treatment used, among other types of recommendations. Recommendations for options (e.g., different tools to use), recommendations for how to make this treatment option (e.g., how to activate the tools created before and / or during the execution of treatment), or recommendations for no treatment Certain treatment recommendations can be made.

  Examples of devices and processes for providing feedback to a clinician during diagnosis and / or treatment of a lesion, including providing treatment recommendations during diagnosis and / or treatment are provided above. In some embodiments, in addition to, or in place of, providing feedback during such diagnosis and / or treatment, the medical device may be configured to perform diagnostic and / or treatment after operation of the medical device in diagnosis / treatment. And / or may be configured to present information about the treatment to a clinician. FIG. 16 illustrates an example of such a process.

  The process 1600 begins at block 1602, block 1604, where the medical device operates to create information regarding the characteristics of the lesion and performance of the treatment, as well as recommendations on how to perform the treatment. The operation of blocks 1602 and 1604 may be similar to the data creation example described above.

  At block 1606, after the treatment, the information created at blocks 1602, 1604 is used by the chronicle creation facility to create a treatment chronicle. The treatment chronicle relates to how the device worked over time, what features of the lesion were detected, what recommendations were made by the medical device, and whether these recommendations were followed by the clinician. May contain information. If treatment detects an error, such as the loss of some or all of the lesions that result in the formation of an embolism or subsequent treatment, the chronicle-making facility analyzes this error to determine the cause of the error. Can decide. For example, if a sensor detects that a portion of the lesion has left the stented clot collector and another sensor has detected the sudden application of force to the stented clot collector just before, a chronicle is created. The facility notes this to the chronicle. The force applied to the stent clot collector exceeds the recommended maximum force recommendation from the medical device, or the medical device is in any other way that is inconsistent with the treatment recommendation. When activated, this can be noted in the chronicle. If such information is included in the chronicle, recommendations on how to avoid this error in future approaches may be made to the clinician.

  Further, in some embodiments, the chronicle creation facility may include detailed information in the chronicle about the lesion and potential causes of the lesion to assist a clinician in diagnosing the lesion. For example, in some embodiments, brief features of the lesion may be output during the treatment (eg, the lesion is viscous), but more detailed information about the composition may be output in the chronicle (eg, the lesion Is mainly composed of cholesterol). In addition, the chronicle making facility may be used to determine whether the lesion was a thrombosis that developed at the site of the lesion, or an embolism that had settled at the site of the lesion, for example, as a result of the injury, in terms of the location of the lesion of interest. The composition can be analyzed. For example, if the lesion is composed primarily of tissues such as smooth muscle cells or atheroma, the lesion may have become a tumor at that site after injury. Another example showed that the composition of the lesion was formed in the area of the body structure with high shear stress, but if the lesion is located in the area of the body structure with low shear stress, May indicate that the lesion was an embolism that had adhered to this site.

  Once the chronicles are created in block 1606, the chronicles are output for presentation to the user (eg, output via a display or stored in memory or transmitted over a network) and a process is performed. 1600 ends.

  In addition, the inventors have determined that the accuracy of the techniques described herein, eg, the accuracy of diagnosis and / or the certainty in which treatment is recommended for intervention of a particular type of pathology, in the model Data used for training and data collected on lesions and used to diagnose lesions or determine how to treat lesions increase as the certainty of responding to lesions and not to other tissues or anatomy increases I understand that.

  In addition, the present inventors have recognized and often found that an insertable or implantable device may have much contact with the correct lesion of interest or other non-anatomy during collection of a measurement. I understand. Otherwise, this is often the case when the probe, including one or more sensors, can make full or partial contact with other anatomy located adjacent or proximal to the anatomy of interest. possible. For example, if the insertable device is manipulated through the animal's vasculature until it comes into contact with a lesion in a blood vessel and then activates to collect a measurement of the lesion, the sensor of the probe may contact the lesion. In addition, there is certainly the possibility that it may come into contact with the vessel wall. As a specific example, rather than puncturing or otherwise contacting the lesion only, the insertable device may be such that some sensors contact the lesion and some contact the vessel wall. First, it can be placed between the lesion and the vessel wall.

  If measurements are collected on one or more other anatomy in addition to the anatomy of interest, such measurements may be used to identify or characterize the anatomy or to determine appropriate treatment for the anatomy. May affect the accuracy of the techniques described herein.

  Accordingly, the present inventors have developed an approach for filtering EIS measurements collected by the methods described herein to remove measurements that are not related to the lesion or other anatomy of interest. More specifically, the inventors have found that filtering data collected to remove or at least reduce the number of measurements that are not associated with a lesion accurately characterizes the characteristics of the lesion and / or Can significantly increase the properties of the model that provide appropriate recommendations for the treatment of. The filtered data (with or without reducing the measurements corresponding to other structures) may also be used to train the model in any of the ways described above.

  In addition, the present inventors have recognized and understood that there are various techniques for filtering data to remove extraneous or outliers, and techniques that utilize machine learning to perform the filtering. Recognize and understand the benefits provided by In such a machine learning process, EIS measurements on the anatomy where the model "passes" the measurements, as well as the anatomy of interest, may be located in regions of the animal's body where the model "filters" the measurements. The model may be trained with EIS measurements on one or more other anatomy. One or more other anatomical structures may be included in the expected (predicted) body structure in the body of various animals in the area to be investigated, for example, the area where a lesion is found or predicted to be investigated. And can be identified. For example, once an area of the animal's body is identified and the anatomy to be measured or treated is identified, it can be found in that area of the animal's body during the setting phase, adjacent or near the anatomy of interest. One or more other anatomical structures that can be positions are identified. The in vitro measurements are then made in vitro with respect to the anatomy of interest (eg, a particular type of lesion, eg, a range of lesion types with a range of features / compositions), and / or one or more other anatomy. Can be recovered. The model may then be trained on this region of the animal's body based on measurements to distinguish between the target anatomy and one or more other anatomy.

  FIG. 17A provides for training a model to characterize a lesion and / or to provide a treatment recommendation with an improved degree of certainty or accuracy, via filtering measurements on anatomy other than the lesion. 5 is a flowchart illustrating a process. Process 1700 begins at block 1702, where training data is received. Training data may include measurements obtained by sampling biological material using sensors of the type described above. Training data may be obtained using in vitro or in vivo techniques. Training data may include measurements related to the pathology to be treated and other anatomy. Also, the identity of the structure to which the measurement corresponds may be entered in training to assist the model in learning to distinguish between the measurement corresponding to the target anatomy and other measurements.

  The training at block 1702 is performed to train the model to distinguish between EIS measurements for the anatomy of interest (eg, a particular type of lesion) and EIS measurements not for this anatomy. obtain. Thus, such a model may divide these EIS measurements into one of two categories: a category “related” to the target anatomy and a category “not related” to the target anatomy. In other embodiments, rather than training the model only to distinguish between these two categories, the model may be trained with EIS measurements for each of a plurality of different anatomical structures, and the input EIS The measurements can be trained to classify the measurements into one of these categories and to ensure that each EIS measurement identifies the corresponding anatomy most reliably.

  The model can be trained at block 1702, for each particular EIS measurement, to identify whether the EIS measurement corresponds to the anatomy of interest and / or the corresponding structure to the EIS measurement. Thus, the model in block 1702 may not be trained to identify or filter on the set of EIS measurements, but rather process the set of EIS measurements to filter the individual EIS measurements in this set. I can do it. Such a set of EIS measurements may include the insertion of the implantable or implantable device at a particular time, such as the collection of EIS measurements at the time of using multiple sensors of the implantable device or the insertable device. It may include measurements from actuation. Distinguishing between measurements in this way can be based on the EIS measurements collected over a specific time or time interval by a sensor in contact with the lesion or other anatomy of interest and the May be able to distinguish between EIS measurements collected by another sensor of the device in contact with another anatomy.

  Once the model is trained to filter in this manner, the filter may be part of a process for training the model to identify and / or classify anatomy (eg, lesions) based on EIS measurements. Can be used with training data. More specifically, at block 1704, the training data is filtered to remove or at least reduce data that is not relevant to the anatomy of interest, such as a particular type of lesion. At block 1708, the filtered training data is used to recognize a relationship between the filtered data and the features of the lesion (eg, according to the embodiments described above in connection with FIGS. 15A-15B), or , The model may be trained to directly recognize the relationship between the filtered data and the treatment recommendations (according to some embodiments described above in connection with FIG. 15C). Block 1708 may be implemented using any of the processes described above.

  In embodiments where the filter of block 1704 is trained to learn which subset of data corresponds to the lesion and which data does not, the process 1700 may include a multi-step training model. It should be understood that one training step is performed at block 1704 and a second training step is performed at block 1706. In some embodiments, the filter of block 1704 and the model of block 1706 are trained using the same data simultaneously, ie, as a one-variable problem. However, in other embodiments, the filter of block 1704 and the model of block 1706 are trained using different data and / or at different times.

  In some embodiments, the training of FIG. 17A may be performed by the techniques described above in connection with FIG. As is evident from the above, the process of FIG. 22 may be used to train one or more models and / or to analyze data in connection with the models to use in discriminating between tissues so that EIS measurements may be made. And an iterative approach to identifying the frequency at which to collect the characteristics extracted from the EIS measurement. At each iteration of the iterative approach, different frequencies / different characteristics are used for training the model to determine whether the model trained on this input can meet one or more performance goals. , Can be selected. In some embodiments, such an iterative process may also be used to train filter models to filter EIS measurements that do not correspond to the anatomy of interest, and to identify and / or characterize the anatomy of interest. Can be used to train a model. In some such embodiments, the EIS measurement is found in the anatomy of interest (eg, a particular type of lesion) and / or in a region of the animal's body that is adjacent or proximal to the anatomy of interest. With respect to the other anatomy obtained, it can be recovered in vitro and / or in vivo. EIS measurements can involve a wide range of frequencies. While training the process according to FIGS. 22 and 17A, an iterative process is used, in which during each iteration a subset of the frequencies is selected and the EIS measurements for these frequencies are determined by two models (the filter model and the biological model). Model for identifying / characterizing structures). Additionally or alternatively, as discussed in connection with FIG. 22, at each iteration, a set of properties may be selected, which may include properties that are present in the EIS measurement and / or properties that result from the EIS measurement. The iterative process involves selecting frequencies and / or characteristics and training based on the selected frequencies and / or characteristics until the process for identifying and / or characterizing the anatomy at each iteration meets one or more performance goals. May continue. Such performance goals may be to reach or exceed the desired accuracy in identifying or characterizing the anatomy, as discussed in connection with FIG.

  After such training, the insertable device and / or the implantable device may be configured to collect EIS measurements at the identified frequency and filter the EIS measurements to not correspond to the desired anatomy. The trained model can be configured to remove the measurement and to identify / characterize the anatomy using the filtered EIS measurement. As noted above, those skilled in the art will understand that applying a trained model may include performing computations using the weights or other values created during training. It is. Thus, the insertable device and / or the implantable device may also be configured with the training determined values so that the device may apply the trained model.

  In some embodiments, different filters may be created for different parts of an animal body (eg, a mammalian body, including a human or non-human body). For example, one filter may be created to distinguish between cardiac tissue (such as a coronary artery wall) and a cardiac lesion (such as a coronary artery clot), and another filter may be created to differentiate between brain tissue (such as the cerebrum). (E.g., the inner wall of a vein) and a lesion that may be found in the brain. Also, different models can be trained on different target anatomy, such as different types of lesions. Thus, different models may include models that are trained with respect to a particular anatomy of interest and with respect to a particular region of the animal's body where the anatomy of interest is located and measured. We have that having different filters for different body parts and / or different anatomy of interest substantially limits the amount of data training each filter than one filter for the whole body, It is understood that the computer processing involved in the learning process can be reduced. Also, different models may improve the accuracy of the filter. However, in some situations, one filter for the anatomy of interest may be used for multiple body parts or the entire body, so all embodiments need to utilize multiple filters for different body parts. It should be noted that this is not the case.

  FIG. 17B illustrates a process that may be used to create a filter for use in block 1704 of FIG. 17A. Process 1720 begins at block 1722, where a body part is identified. Examples of body parts include, but are not limited to, the heart or a particular part thereof, the brain or a particular part thereof, the liver or a particular part thereof, the kidney or a particular part thereof, the veins of the limbs, and the like. After the body part is identified, one or more biological materials (including tissue and / or lesions) of the body part may be identified at block 1724. For example, if a heart is identified at block 1722, certain tissues of the coronary arteries (eg, smooth muscle cells, elastic fibers, outer elastic lamina, internal elastic lamina, loose connective tissue, and / or endothelial cells), and / or Blood clots, which can generally be found inside the coronary arteries, can be identified at block 1724.

  At block 1726, data related to the biological material identified at block 1724 may be collected. This data may represent a measurement obtained by sampling the biological material, which in some embodiments is performed using a sensor mounted on an invasive probe. In some embodiments, the data may include a measurement of a spectrum, such as a collection of EIS samples obtained at different frequencies, associated with a particular biological material of the body part.

  At block 1728, the collected data is labeled so that the data represents the particular tissue or lesion to which it corresponds. For example, data corresponding to measurements obtained from coronary artery endothelial cells may be labeled "coronary endothelial cells"; data corresponding to measurements obtained from coronary artery elastic fibers may be labeled "coronary elastic fibers". The data corresponding to the measurement obtained from the thrombus found in the coronary artery may be labeled "thrombus in the coronary artery"; the data corresponding to the measurement obtained from the fibrin found in the coronary artery is " Fibrin in coronary arteries ".

  At block 1730, a filter may be trained using the labeled data. Specifically, the filter is designed to distinguish between lesions (eg, thrombus, fibrin, or other types of blood clots) and data associated with tissues (endothelial cells, elastic fibers, loose connective tissue, etc.). Get trained.

  While the techniques of FIGS. 17A-B have been described above in connection with filtering measurements on other anatomy, it should be understood that the techniques of filtering are not so limited. In other embodiments, the filtering techniques described in connection with FIGS. 17A-17B may additionally or alternatively be used to identify or filter erroneous measurements. Erroneous measurements may result from any potential sources of error and may include EIS measurements with magnitude and / or phase values that are inaccurate. An erroneous measurement may affect the accuracy of the same or similar method as a measurement corresponding to other anatomy. Thus, to train the model to distinguish between erroneous and erroneous measurements, known erroneous measurements, along with indicators that the measurements are erroneous, along with non-erroneous measurements, An input training process may be performed. The trained model, which may also be trained to distinguish between measurements on the anatomy of interest and measurements on other anatomy, can then be used by the device to filter the measurements. .

  We understand that conventional techniques for identifying the location of an insertable device relative to a lesion in an animal's body are often insufficient. For example, one conventional technique for locating blood clots in the vasculature involves the use of x-ray images associated with x-ray reflective probes. Specifically, when an invasive probe having a portion that reflects x-rays is inserted into an animal's tract, the position of the probe can be monitored using x-ray imaging. Unfortunately, clots usually do not reflect x-rays; thus, x-ray images do not provide an indication of clot positioning. As a result, the process of contacting a clot with an invasive probe using it is often cumbersome and necessarily involves a great deal of effort. This process may unnecessarily increase the duration of the operation, or worse, may result in damage to the inner wall of the vessel as the clinician repeatedly manipulates the probe in the search for a clot.

  In this regard, the present inventors have recognized the advantages of the technique for determining whether a probe is in contact with a lesion. Thus, some embodiments provide that the investigated anatomy is a anatomy of interest, such as a lesion (eg, a blood clot), or another anatomy, such as the tissue in which the lesion is formed (eg, the inner wall of a vessel). ). In some embodiments, machine learning techniques may be used to identify whether the anatomy is the anatomy of interest. Thus, the model can be trained using known data identifying measurements associated with different anatomy. After training, the model may be able to distinguish between anatomical structures. In some embodiments, the process of FIG. 17B is used to train a model, but alternative processes are possible.

  FIG. 17C illustrates a process that may be used to assist a clinician in guiding an invasive probe. The process 1740 begins at block 1742, where data is received, for example, at a processor of a medical device, from one or more sensors mounted on an invasive probe inserted into an animal (eg, the vasculature of the animal). Is done. The received data may include, at least in some embodiments, impedance measurements associated with one or more anatomy of the animal.

  In some embodiments, the data received at block 1742 may be multiple sets of sensor readings collected over a period of time. During this period, the insertable or implantable device may not have to be moved so that in some cases all of the sensors of the device are in contact with the same substance over the set of measurements. Multiple measurements in such cases may assist in producing a more reliable result in the identification or characterization of the anatomy contacted by the sensor of the device. In some embodiments, for example, a model trained to identify or characterize a anatomy is one in which the sensor readings correspond to a particular anatomy or a particular feature (eg, composition) of the anatomy. A value representing certainty may be created. This model may receive additional data as input, which may adjust for higher or lower certainty. Thus, collecting a plurality of data over a period of time may help to ensure that the anatomy is or is not the desired anatomy. In some such embodiments, for example, the anatomy contacted by the device may be such that at least three spectra are obtained in less than 3 seconds, at least 5 spectra are obtained in less than 3 seconds, or at least 10 The spectrum may be sampled so that it is obtained in less than 3 seconds.

  At block 1744, it may be determined whether the measurement corresponds to a target anatomy, such as a particular type of lesion. In some embodiments, this determination may be made in part using a machine learning filter trained by process 1720 (FIG. 17B), thereby training to distinguish between anatomy. obtain.

  If the EIS measurement is determined to correspond to another anatomy that is not the structure of interest, then at block 1746, information is output to the user, indicating that the probe has not yet contacted the anatomy of interest. May be notified. This information may be output in any suitable way, for example using a display or an audio device. Based on this information, the clinician may decide to continue guiding the invasive probe in the animal, and the process 1740 may repeat via block 1742, block 1744, and block 1746.

In some embodiments, the determining step of block 1744 may relate to whether all EIS measurements collected by the device at a particular time or over a particular time period correspond to the anatomy of interest. For example, the determination may be whether the device is in complete contact with the structure of interest (eg, a lesion), or whether one or more sensors are in contact with another biological material. If not all sensors are determined to be in contact with the structure of interest, then at block 1746, information is output indicating that the device is not in complete contact with the anatomy of interest (eg, a lesion). obtain. In another embodiment, the determining step of block 1744 is to identify which device's sensor (if any) is generating a read indicating that the sensor is in contact with the anatomy of interest. Can be done. Distinguishing between the sensors in this manner can assist the clinician in operating the device for proper and complete contact with the target anatomy. For example, if the sensor is positioned longitudinally along the length of the insertable device, so that the device subsequently approaches the target anatomy and then contacts the target anatomy, the sensor will have the desired structure May be generated over time, and then a sensor at the distal end may generate a value indicating that the sensor is in contact with the structure, and then the device is properly positioned. After that, a value may be generated that indicates that all sensors are in contact with the structure. If the clinician moves the device too far, the distal sensor may generate a value indicating that the sensor is not in contact with the desired structure.

  If it is determined that the investigated biological material is the desired anatomy, then information may be output at block 1750 for this effect. The clinician may then perform the following steps: actuate the invasive probe to determine one or more characteristics of the lesion (described in connection with block 104 of FIG. 1); create a treatment recommendation based on the characteristics of the lesion and Activating the medical device to output (described in connection with block 106 of FIG. 1); selecting a treatment option based on treatment recommendations (described in connection with block 108 of FIG. 1); and / or selecting the selected treatment. Any of the treatments of the lesion using the options (described in connection with block 110 of FIG. 1) may be performed.

Examples The following describes examples of various scenarios in which medical devices and technologies may be used. However, it should be understood that the embodiments are not limited to operation according to any one of these examples.

Example 1
One example of how the techniques described herein may be used employs an invasive smart guide wire. This invasive guidewire can be used for manipulating the vasculature. Using the sensors and analysis techniques described herein, an invasive guidewire may characterize the tissue / material in contact with the guidewire and convey this tissue / material characteristic to the clinician. Also, an invasive guidewire may assist additional devices to reach the intervention site in the patient.

  In this embodiment, the guidewire includes a sensor (preferably an EIS sensor), an impedance spectrometer, and a handle. Also, the guidewire can include additional components that can be inserted along its length during use. The sensor can be used to detect and characterize the nature of the tissue / material that contacts the sensor. For example, the sensor may be used to determine tissue / substance composition when used with an impedance spectrometer to make high frequency impedance measurements. Both the sensor and the impedance spectrometer are preferentially located at the invasive tip of the guidewire, allowing tissue adjacent to this tip to be featured without the need for long wires connecting the sensor to the impedance spectrometer It can be attached. This design may reduce electrical noise that may otherwise be inserted into the electrical signal when the impedance spectrometer is located outside the object.

  The handle may include additional components, such as components for communicating with a user, recording and transmitting data during and after surgery, processing data, and powering the device. Examples of such components include a user-readable feedback portion such as a display or indicator light, components for transmitting data wirelessly or via a cable, a database, a processor, and a battery. The handle can be detached from other device components and can be removably connected to the guidewire's own circuitry.

Example 2
The guidewire described in Example 1 can be used by a clinician to determine the optimal treatment strategy for a patient with an inhibited artery. The clinician can use the guidewire to characterize the tissue / material blocking the artery and then use this information to choose between different potential treatments. In some embodiments, the guidewire provides a clinician with a treatment recommendation based on one or more characterizations made in advance, and optionally based on data from previous treatments made with this guidewire. Can provide.

  In this example, a clinician can use a guidewire to assess and treat arterial lesions. The clinician can begin by manipulating the guidewire, optionally using the handle, to the site of the thrombus, and then piercing the thrombus. The clinician may then use the guidewire to make a measurement of the composition of the thrombus and / or tissue / material blocking the artery. The clinician can then determine the optimal treatment for the blocked artery based on this measurement. For example, a clinician may decide to use a stent device if the inhibiting tissue is made up of cells from the patient's arterial wall. If the inhibiting tissue is a thrombus, the clinician instead decides to measure its viscoelastic properties, and then uses a suction catheter or stent to remove the clot based on this information You can decide if you should.

  In some embodiments, the clinician may also receive treatment recommendations from a guidewire. The treatment recommendations can be based on characterization performed by the guidewire on the lesion of the artery and / or based on data collected during previous use of the guidewire.

  After the end of treatment, the clinician may remove the handle from the guidewire and insert the appropriate interventional device with the guidewire.

Example 3
A further example of a device that may be used in accordance with the techniques described herein is a smart stent-stent-retriever. The stent thrombus collector can be used to collect a clot from a patient. By using the sensors and analysis techniques described herein, an invasive stented thrombus collector can characterize the clot in contact and communicate this tissue / material characteristic to the clinician.

  In this embodiment, the stent thrombectomy device includes at least one sensor (preferably at least one EIS sensor and / or EIT sensor), a measurement unit, and a handle. Stent clot collectors may include multiple sensors at multiple strategic locations so that information about the clot that is in contact can be obtained from multiple locations in the clot. If the stent clot collector includes more than one sensor, the sensors may be able to detect different properties of the clot that it contacts. For example, a stent clot collector may include one or more sensors capable of detecting the integration of the clot with the clot, one or more sensors capable of detecting the position of the stent clot as a function of time, and / or Or it may include one or more sensors capable of sensing the force applied to the clot. The integration of the clot with the stent clot collector can be determined by sensing the stent inductance and / or EIT signal as a function of time. Since the values of stent inductance and EIT vary with the stent's spread and surrounding environment, constant values of these properties will indicate that the stent has reached maximum spread and integration into the clot. . Motion sensors can be used to detect the position of the stent clot collector as a function of time. This property allows the clinician to understand the movement of the stent clot collector in the patient and determine the number of passes that the stent clot has made during clot collection. Also, a stress sensor may be included to measure the force applied by the stent clot collector on the clot or tissue / material.

  The measurement section of the stent clot collector may be an impedance spectrometer and / or a tomography section. This part is preferentially located near the tip of the stent clot collector, which requires a long wire to connect the clot adjacent to the stent clot collector to the sensor at the measurement site Can be characterized without This design may reduce electrical noise that may otherwise be inserted into the electrical signal when the impedance spectrometer is located outside the object.

  The handle may include the additional components described in Example 1. It can also include a robotized pulling mechanism to allow accurate and automated clot collection.

Example 4
The guidewire described in Example 1 and the stent-type thrombectomy device described in Example 3 are used jointly by clinicians to determine and execute optimal treatment strategies for patients with blocked arteries Can be done. A clinician can use a guidewire to characterize the tissue / material blocking the artery, and then use a stented thrombectomy device to collect clots and / or thrombi. Optionally, data can be collected during clot collection and uploaded to a database for subsequent analysis.

  In this embodiment, a clinician may use a combination of smart devices to treat a patient with an inhibited artery. The clinician may begin the evaluation of the lesion described in Example 2 by inserting a guidewire with a sheath and using the guidewire (with the invasive probe described above). If, based on the information and / or recommendations provided by the guidewire, the clinician then decides to use a stent clot collector, the clinician removes the guidewire and places the sheath in place. Leave and insert a stent thrombus collector along the sheath to guide the thrombus collector to a clot and / or thrombus. After the stent pierces the clot and / or thrombus, a sensor incorporated in the stent-type thrombus collector can detect the form of the clot and / or thrombus and provide this information to the clinician as a function of time. (Eg, on an external display). For example, an EIS sensor and / or EIT sensor can characterize the integration of a clot and / or thrombus with a stent, and the shape and composition of the clot and / or thrombus. Stent clot collectors may also use data from previous clot and / or thrombus collections to provide treatment recommendations to clinicians. Therapeutic recommendations include, for example, a signal that the integration of a clot and / or thrombus with a stented thrombus collector is optimal, and / or recommendations regarding the appropriate speed and force to withdraw the clot and / or thrombus. be able to.

  At this point, the clinician may act based on the information and / or recommendations provided by the stent-stent-retrieve to retrieve the clot and / or thrombus. A clinician may decide to use an automatic withdrawal mechanism built into the stent clot collector to collect the clot. The automatic withdrawal mechanism then uses the speed and force determined by the stent clot collector based on data received from a previous clot and / or thrombus collection database to attain a rate of clot at a certain rate. And / or the thrombus may be withdrawn. If the clot and / or thrombus move away from the stent clot collector, the stent clot may signal the clinician using an alarm. The clinician may then pierce the clot and / or thrombus again and start the retrieval process again.

  At the end of clot and / or thrombus collection, all data collected during the intervention can be transmitted to a database for later analysis.

Example 5
Another example of a device that may be used in accordance with the techniques described herein is a smart suction catheter. This suction catheter can be used to retrieve a clot from a patient. By using the sensors and analysis techniques described herein, an invasive suction catheter can characterize the clot in contact and communicate this tissue / material characteristic to the clinician.

  In this embodiment, the suction catheter includes at least one sensor (preferably, at least one EIS sensor and / or EIT sensor), a measurement unit, and a handle. As in Example 3, the suction catheter may include multiple sensors at multiple strategic locations so that information about the clot that is in contact can be obtained from multiple locations within the clot. If the aspiration catheter includes more than one sensor, the sensors may be able to detect different properties of the clot in contact. For example, the suction catheter may be one or more of the sensors described in Example 3 (i.e., one or more sensors capable of detecting the integration of the blood clot with the suction catheter, one or more sensors capable of detecting the position of the suction catheter as a function of time). Sensors and / or one or more sensors capable of sensing the force applied to the clot). Also, the suction catheter may include additional sensors that can monitor blood flow in the suction catheter.

  The measuring part and the handle of the suction catheter are the same as the measuring part and the handle of the stent-type thrombus collector described in the third embodiment.

Example 6
The guidewire described in Example 1 and the suction catheter described in Example 5 can be used cooperatively by clinicians to determine and execute optimal treatment strategies for patients with blocked arteries. A clinician can use a guidewire to characterize the tissue / material blocking the artery and then use a suction catheter to retrieve the clot and / or thrombus. Optionally, data can be collected during clot collection and uploaded to a database for subsequent analysis.

  In this embodiment, a clinician may use a combination of smart devices to treat a patient with an inhibited artery. The clinician may begin evaluation of the lesion as described in Example 2 by inserting a guidewire and using the guidewire. If the clinician decides to use the suction catheter next based on the information and / or recommendations provided by the guidewire, the clinician then inserts the suction catheter along the guidewire and inserts it. Induce a clot and / or thrombus and initiate the aspiration process. During aspiration of the clot and / or thrombus, an external display will show the progress of the removal, the shape and composition of the clot and / or thrombus as sensed by the EIS and / or EIT sensors, and the clot via the aspiration catheter. And / or provide information to the clinician regarding passage of the thrombus. The smart suction catheter may also determine the optimal time to initiate clot and / or thrombus removal based on the integration of the clot and the suction catheter, and signal this condition to the clinician. The clinician may then begin removing the clot and / or thrombus. If the clot and / or thrombus is remote from the suction catheter, the suction catheter may signal to the clinician using an alarm. The clinician may then pierce the clot and / or thrombus again and start the retrieval process again. If the sensor detects that the thrombus has been completely aspirated and passed along the tube of the inhaler, another message indicating successful removal may be created and output.

  At the end of clot and / or thrombus collection, all data collected during the intervention can be transmitted to a database for later analysis.

Example 7
The guidewire described in Example 1 can be used to treat patients with chronic total occlusion (CTO). In this case, the patient's arteries are obstructed by an old, hard thrombus, which can be difficult for the clinician to pierce to re-establish blood flow. A clinician may sense the location of the lesion and use a smart guidewire through the lesion. During operation, the guidewire can provide the clinician with information as to when puncturing of the lesion will be initiated and when the guidewire will travel through the lesion into the lumen of the artery. If the thrombus is too hard to pierce, the clinician can instead pass a guidewire through the arterial wall adjacent to the lesion. In this case, the guidewire can provide the clinician with continuous information regarding its location within the atheroma / plaque. This may assist a clinician in avoiding a puncture of a blood vessel.

Example 8
The guidewire described in Example 1 can be used by clinicians in diagnosing and / or treating peripheral medical conditions. Examples of peripheral pathologies include a thrombus formed in a deep vein or artery, or a thrombus formed in an artificial vein or artery. Guidewires can be used to determine the optimal treatment strategy for patients with peripheral medical conditions. The clinician can use the guidewire to characterize the tissue / substance that is obstructing the vessel, and then select among different potential treatments based on this information. In some embodiments, the guidewire is based on one or more characterizations made by the guidewire and, optionally, based on data from a previous treatment performed by the guidewire, and provides a clinician with a treatment recommendation. Can be provided.

Example 9
As a further example, any one of the invasive probes described above can be used to estimate the age of a clot (eg, thrombus). The age of the clot (ie, the time the clot has been formed) can be determined based on one or more characteristics of the clot, such as the composition of the clot. Different treatments or combinations of treatments may be provided based on the age of the clot as determined from these characteristics. For example, if the clot is less than 14 days, one treatment may be recommended, and if the clot is more than 14 days, a different treatment may be recommended.

  Additionally or alternatively, at least some of the devices and techniques described herein can be used to identify whether a anatomy is a healthy tissue. For example, the device / technique can be used to determine if the vessel wall is healthy or if atherosclerotic plaque or calcification has formed in the vessel wall. In such cases, the anatomy with which one of the devices described herein is in contact may be a vascular wall or an atherosclerotic plaque (or other lesion) and the techniques described herein may be used. It can be used to determine if it is one of these anatomy. Based on this identification, different treatment recommendations can be provided.

Methods of Operating Medical Devices for Use in Oncology The present inventors have recognized and appreciated that conventional techniques for the testing of potentially cancerous cells are often inadequate. For example, one of the conventional techniques for testing potentially cancerous cells uses a needle to remove a tissue sample. Conventional imaging techniques such as x-ray, ultrasound, or nuclear magnetic resonance imaging (MRI) have been used to assist the clinician in manipulating the needle insertion. However, images created using these techniques are often inaccurate or blurry, so that the clinician can determine whether the needle is in contact with the cell or tissue of interest. Making it difficult. As a result, the diagnosis and / or treatment of cancerous cells using such techniques is often inaccurate. As a result, when trying to determine whether a particular lesion is cancerous, a needle intended to test a potentially cancerous lesion does not actually come into contact with the lesion, but instead There is a significant risk of contact with healthy tissue, leading to incorrect samples and incorrect medical conclusions. Similarly, when trying to remove cancerous cells, two undesirable situations can occur: healthy tissue can be removed with the cancerous cells, or some cancerous cells can remain unremoved. obtain.

  Thus, according to some embodiments described herein, the medical device can include the presence of cancerous cells / tissues, the characteristics of cancerous cells / tissues, and / or the type of cancerous cells / tissues (eg, , Cell tumors, lymphomas, myeloma, neoplasms, melanomas, metastases, or sarcomas). For example, the machine learning techniques described above can be used to distinguish between cancerous cells / tissues and non-cancerous biological materials and / or to characterize cancerous cells / tissues. Further, techniques of the type described herein, including machine learning techniques, provide recommendations for methods of treating cancerous cells / tissues based at least in part on the characteristics of the cancerous cells / tissues. obtain. For example, ablation or removal of cancerous cells / tissues, and methods of ablation or removal may be recommended in some situations.

  Examples of medical devices, sensors, and methods for detecting tissue / substances of cancerous cells are detailed above with respect to FIGS. Examples of techniques that the medical device can perform and / or that the medical device can operate to perform are described below with reference to FIGS.

  FIG. 31 illustrates an exemplary process 3100 that may be performed by a medical device operating according to some of the techniques described herein. In the example of FIG. 31, the sensors may be located on a diagnostic and / or therapeutic device such as a needle, ablation catheter, radio frequency probe, robotic probe, laparoscopic, or cutting device. In some embodiments, the sensor is located near a distal end of the medical device. The medical device may make a treatment recommendation based on the characteristics of the cancerous cells determined using the sensor. It should be understood that the processes described herein are not limited to use with invasive probes. In some embodiments, the techniques described herein may not be designed for use within an animal's body or only for use within an animal's body, and / or Can be used with systems and devices that include non-invasive probes that can be designed for use with anatomy that includes tissue outside the body. For example, in some embodiments, the devices, systems, and techniques described herein may be used for the diagnosis and / or treatment of surface lesions, such as skin cancer or other skin conditions.

  Process 3100 begins at block 3102, where the invasive probe of the medical device may be one or more features (eg, size and / or size) of a lesion proximal to the sensor, which may be a cancerous tissue or cell. (Composition). Prior to the start of process 3100, an invasive probe may be inserted into the animal's body and moved proximal to the predicted lesion location. Next, the medical device is activated to detect when the sensor contacts the lesion. Contact of a lesion or tissue that is known to be or is potentially cancerous can be determined by assessing changes in the value output by the sensor over time (eg, changes in impedance). , Or by using the machine learning techniques described in connection with FIG. 17C. For example, a medical device may require that the sensor of an invasive probe not be in contact with cancerous tissue / cells or in contact with a type of tissue of which the lesion is known to be part. One result may be output (eg, to a user via a user interface).

  For example, when examining a lesion by moving an invasive probe through the animal toward the lesion, the medical device may output a value representative of the tissue in contact. This value may, in some embodiments, be a quantitative value, such as a yes / no or true / false indication of whether the invasive probe is in contact with the lesion. Contains the original value.

  The medical device analyzes the biological material contacted by the invasive probe, including the tissue contacted by the invasive probe, and identifies any biological material for which the invasive probe is "abnormal" and thus may be part of the lesion. By determining whether it is in contact, it can be determined whether the invasive probe is in contact with the lesion. The medical device may, in some embodiments, determine whether the biological material contacted by the probe is “abnormal” by assessing the position of the invasive probe in the animal, It may represent a biological material that the probe may be expected to contact.

  Additionally or alternatively, the medical device may determine whether the invasive probe is in contact with the lesion based on predictions about the lesion, which may be input by the clinician as a result of the preliminary diagnosis. For example, clinicians may provide information that preliminarily characterizes the lesion, for example, whether the lesion is a vascular or organ lesion, or if an organ lesion, what organ, A prediction of the composition or state of the diseased tissue or cells (eg, not healthy, inflammatory, cancerous, in a diseased state, etc.) can be entered. In embodiments where such information is entered, the clinician may enter information that preliminarily characterizes the lesion individually, or may be associated with information that preliminarily characterizes such lesions. A selection of a preliminary diagnosis of the lesion may be made (eg, by selecting a particular class of atheroma, other information, such as the predicted atheroma composition and its location in the vasculature, May be selected as well). If the invasive probe moves through the animal, the medical device compares the biological material contacted by the invasive probe with the preliminary characterization of the lesion to determine whether the invasive probe is in contact with the lesion I can do it. For example, if the lesion has been pre-diagnosed as a brain lesion that may be a brain tumor, the medical device may determine whether the invasive probe has contacted abnormal brain tissue and / or whether the invasive probe has cancerous status. It can be determined whether or not it is in contact with brain tissue and the result can be output.

  In other embodiments, rather than simply providing a dual value that indicates whether an invasive probe is in contact with a lesion, the medical device may fluctuate, for example, as the probe moves through the body. The sensor of the invasive probe may output a value representing the identity, quantity, and / or relative abundance of the biological material or biological materials with which it is in contact. The value representing the one or more substances may be used to identify a substance, such as a listing of substances identified from impedance spectra determined using the techniques described herein (including the machine learning techniques described above). Can be This value may, in other embodiments, be a numerical value, such as a value detected by the sensor (eg, an impedance value or impedance spectrum) or other value.

  The probe and its sensor may be moved until it makes contact with the lesion, at which point the results output by the medical device may change after the contact is made. In this way, the location of the lesion may be determined using an invasive probe, and a determination can be made that the invasive probe is in contact with the lesion.

  The invasive probe may further be optionally operated to determine the geometry of the lesion. For example, the geometry of a lesion that potentially contains cancerous tissue (eg, a tumor) may, in some embodiments, cause the invasive probe to move proximal to the lesion, and when the sensor of the invasive probe Can be determined by identifying whether or not they are in contact with. For example, if analysis of the values output by the invasive probe determines that the lesion includes cancerous tissue, the invasive probe may be moved, and with different sensors over time, individual sensors A determination may be made as to whether contact is made. Next, the amount of movement of the invasive probe (e.g., measured using an accelerometer as described above) and the position of the sensor on the invasive probe are analyzed by the medical device to determine one or more dimensions of the cancerous tissue. The geometry of the cancerous tissue in the animal can be determined, including:

  In some such embodiments, the medical device may determine one or more treatment recommendations for the lesion based on the geometry of the lesion.

  In one treatment protocol that may be implemented in embodiments such as the one illustrated in FIG. 31, ablation may be used as a first option for treating cancerous tissue. Thus, at block 3104, an ablative device, such as a needle or radio frequency probe, is inserted into the animal. In some embodiments, the ablation device may include an invasive probe that includes a sensor of the type described herein. The ablation device may be moved until the ablation device determines that contact with the cancerous cells or tissue has been formed (however, embodiments are not limited to operation with ablation devices that include invasive probes. It should be understood that in other embodiments, the invasive probe is part of another medical device, and after placement of the invasive probe, the ablation device is placed proximal to the invasive probe, and thus (Migrated until placed proximal to cancerous cells / tissue).

  At block 3104, after placing the ablation device proximal to the cancerous cells / tissue, the ablation device operates to ablate the cancerous cells / tissue. After the treatment time, the ablation device may operate to determine whether the ablation device has an effect on the cancerous cells / tissues. For example, in some embodiments, a treatment recommendation may be developed to guide the clinician in performing the ablation, including whether the ablation is effective and whether the ablation should continue. Thus, at block 3106, the sensor may provide information indicating whether the ablation device is still in contact with the cancerous cells or tissue. This determination may be made using the techniques described herein, including the machine learning techniques described above. This information can be used to provide treatment recommendations, such as whether to stop or continue cauterization, or to confirm the placement of the invasive probe before deciding whether to stop cauterization. Can be used.

  In some embodiments, the ablation device may include a plurality of different electrodes used to cauterize, such as different electrodes located at different locations, wherein the different electrodes operate to ablate portions. And may be individually operable so that others do not operate to cauterize at the same time. In some embodiments, each ablation electrode may be located near a sensing electrode, where the sensing electrode is operated by the techniques described herein to determine the biological material that the sensing electrode has contacted. I do. The ablation device may use the sensing electrodes to determine whether a particular portion of the ablation device is in contact with cancerous tissue / cells or non-cancerous tissue / cells. In some such embodiments, in response to determining that a portion of the ablation device is in contact with the non-cancerous tissue, the ablation device limits the ablation to only cancerous tissue, In order to minimize the damage that can be done to the tissue, the activation of some of the ablation electrodes of the ablation device may be stopped or prevented.

  In this manner, the clinician may abort the ablation if the treatment is ineffective and the clinician may proceed with the ablation by continuing the ablation only while the ablation device is in contact with the cancerous cells / tissues. Only sexual tissue can be cauterized. Thus, the clinician may be more confident at the end of the treatment whether the treatment was successful, and if so, that all of the cancerous tissue / cells have been cauterized. In this way, the risk of cauterizing healthy tissue and / or leaving cancerous cells unburned is reduced.

  Thus, as illustrated in FIG. 31, if the ablation device is still determined to be in contact with the cancerous cells / tissue, the process 3100 proceeds to block 3108 where a recommendation to continue the ablation is provided. Provided and iterates to block 3104. On the other hand, if it is determined that the ablation probe is no longer in contact with the cancerous cells / tissue, a recommendation to stop ablation is provided at block 3110. This process can be repeated by repositioning the ablation device. If no more contact with the lesion can be made, even after several attempts to reposition the ablation device, process 3100 ends.

  FIG. 32 illustrates an example of a method of operating a medical device to generate a treatment recommendation for a cancerous cell / tissue according to another embodiment. In the embodiment of FIG. 32, the medical device may include a plurality of sensors located along the outside of the probe, for example, in the example of FIG. 3 described above. As will be apparent from the above, using such a sensor arrangement, several different characteristics of a cancerous lesion, including the composition of the cancerous lesion, can be determined. For example, by performing an EIS process on a cancerous lesion, the composition of the cancerous lesion can be determined as described above. In some embodiments, the trained machine learning models described above may be used to determine the composition or other characteristics of a cancerous lesion.

  The process 3200 of FIG. 32 begins at block 3202, where a medical device is inserted into the body of an animal subject to detect one or more characteristics of a cancerous lesion, eg, the composition of the cancerous lesion. Works as follows. Based on features including this composition, the medical device may select a recommended treatment option at block 3204. Based on this composition, the process 3200 may determine the type of cancerous pathology investigated and provide appropriate treatment recommendations. The medical device may be configured such that the biological material may be identified using the impedance spectrum, such as in other embodiments described above, and the impedance spectrum and other electrical signals for different biological materials may be identified based on the biological material. With information about the characteristic features (eg, effective capacitance), as well as information about the composition of the different lesions. Further, medical devices can be configured with different treatment recommendations for different types of lesions, such as different types of cancerous lesions. In one example, if the impedance spectrum for the different biological materials of the lesion is used to determine that the cancerous lesion is or is a cell tumor, a recommendation to remove the cancerous lesion may be provided. In another example, if the cancerous lesion is determined to be part of melanoma, high frequency ablation may be recommended. The medical device may select a treatment option in any suitable manner.

  After the treatment is recommended at block 3204, the medical device may monitor the performance of the selected treatment option at block 3206. The medical device monitors treatment using one or more sensors, for example, one or more sensors characterized in block 3202 or one or more sensors of a treatment device operative to provide treatment. obtain. For example, if cautery is recommended at block 3204, the clinician may insert a cautery device. The ablation device may include a sensor, for example, a temperature sensor, for detecting a condition of the cancerous lesion when the ablation is being performed. The sensor may determine whether the ablation was successful by determining whether the cancerous lesion was ablated or frozen.

  At block 3028, information regarding the status of the treatment is output by the medical device via the user interface for presentation to the clinician. Thereafter, process 3200 ends.

  One example of monitoring treatment is provided in terms of creating a treatment advisory, but it should be understood that similar techniques could be used to create an error or other message for the clinician regarding the condition of the treatment. It is. For example, if a sensor on the treatment device indicates the presence of a cancerous lesion at a certain time, after which the sensor no longer detects the cancerous lesion, the medical device may indicate that the treatment device is improperly placed or It can be determined that the sexual lesion has disappeared. This may indicate either that the device needs to be repositioned or that the cancerous lesion has moved. A message to the clinician via the user interface may indicate such a potential problem.

  Further, while the example of FIG. 32 describes a method of operating a medical device to provide treatment recommendations related to the selection of a first treatment and to subsequent methods of delivering the treatment, embodiments are not described herein. It should be understood from the above that the present invention is not limited to this. For example, in some embodiments, a medical device may include one or more sensors described herein and relates to a method of operation of the device without making an initial recommendation to use the device. It can be activated to provide a treatment recommendation. For example, a needle or radio frequency probe may include one or more sensors for generating data regarding the status or performance of a therapy, as described above, and may generate a therapy recommendation.

  FIG. 33 illustrates a process 3300 that may be performed by the medical device of some embodiments to create a treatment recommendation.

  Process 3300 begins at block 3302, where the medical device determines one or more characteristics (eg, size and / or composition) of the cancerous lesion using the techniques herein. Activate The medical device may receive one or more features from a component of the medical device. For example, there is one or more sensors included in the medical device and / or another component that creates features based on data generated by the sensors. This feature may include, in some embodiments, the composition of the cancerous lesion. Additionally or alternatively, the characteristic may be a location of the cancerous lesion in the body, one or more dimensions (eg, length, thickness, etc.) of the aggregate of the cancerous lesion, a temperature of the cancerous lesion, or a sensor as described above. Other information that can be determined based on the type of

  At block 3304, the medical device compares the feature received at block 3302 with one or more conditions for one or more treatment options. The medical device may be configured with information regarding a plurality of different available treatment options, each of which may be associated with one or more conditions associated with one or more characteristics of the cancerous lesion. . Treatment options may include cautery, removal, and biopsy. Examples of such conditions relating to the composition of the cancerous lesion are described above in connection with FIG. For example, the trained machine learning model described in connection with FIG. 15B may be used to determine the relationship between the characteristics of the cancerous lesion and options for successful treatment.

  The medical device may compare the characteristics of the cancerous lesion with the conditions for one or more treatment options to determine which conditions are met. In some embodiments, the set of conditions for the treatment options are mutually exclusive such that the cancerous lesion can match only one set of conditions, and thus only one treatment option can be selected. Can be In other embodiments, the set of conditions may not be mutually exclusive, so the medical device identifies the option with the most corresponding condition and the option with the most corresponding condition. This can determine which treatment options are recommended. For example, if different conditions are associated with different ranges of values, such as ranges of impedance spectra, the conditions may be determined by identifying the range where the values for the lesion most closely match. The closest match is, for example, the range where the impedance spectrum or other value of the lesion is within or furthest from the boundary value for the range, or has the most overlap with the range. obtain.

  At block 3306, based on the comparison, the medical device may output a treatment option recommendation via the medical device user interface, and the process 3300 ends.

  One skilled in the art will appreciate that there are many ways to set conditions regarding treatment options that can be used in connection with processes such as process 3300 of FIG. For example, a value for a characteristic of a cancerous lesion to use as a condition for the selection of a treatment option is the value between this value, the type of cancerous cell / tissue, and the correspondence between treatment success with various treatment options. May be hard coded into the medical device after at least some experimentation to determine However, the present inventors have developed a system for learning such relationships and conditions based on, among other information, the characteristics of cancerous cells / tissues and the successful treatment of cancerous cells / tissues. Recognize and understand the benefits of For example, a machine learning process, for example, a machine learning process that may include feature extraction and / or classification, may be implemented in some embodiments.

Computer Implementation Techniques that operate according to the principles described herein may be implemented in any suitable manner. A series of flow charts showing various process steps and acts for characterizing vascular lesions and / or creating one or more treatment recommendations for treating a lesion are included in the above discussion. . The process and decision blocks of the above flowchart represent steps and acts that may be included in an algorithm for performing these various processes. The algorithms resulting from these processes are integrated with one or more single-purpose or multi-purpose processors, and may be implemented as software directed at the operation of the processors, and may include digital signal processing (DSP) circuits or It may be implemented as a functionally equivalent circuit, such as an application specific integrated circuit (ASIC), or may be implemented in any other suitable manner. It should be understood that the flowcharts contained herein are not indicative of the syntax or operation of any particular circuit or any particular programming language, or type of programming language. Rather, these flowcharts are used by those skilled in the art to construct circuits or implement computer software algorithms to perform processing on particular devices that have implemented techniques of the type described herein. The functional information to be obtained is exemplified. Also, unless otherwise stated herein, the particular order of steps and / or acts set forth in each flowchart is merely an example of an algorithm that may be implemented, and the principles described herein may not be used. Implementations and embodiments may vary.

  Thus, in some embodiments, the techniques described herein are implemented as software, including application software, system software, firmware, middleware, embedded code, or any other type of suitable computer code. It can be embodied in computer-executable instructions. Such computer-executable instructions may be written using any of a number of suitable programming languages and / or programming or scripting tools, and are executed on a framework or virtual machine. It can be compiled as executable machine language code or intermediate code.

  When the techniques described herein are embodied as computer-executable instructions, these computer-executable instructions may each include one or more instructions for completing the execution of an algorithm operating with these techniques. It can be done in any suitable way, including many functional facilities that provide the operation of However, the illustrated "functional equipment" is a structural configuration of a computer system that integrates with one or more computers and, when executed by the computers, causes one or more computers to perform specific operational roles. Element. Functional equipment may be part or all of a software element. For example, a functional facility may be implemented as a function of a process, or as another process, or as any other suitable unit of processing. Where the techniques described herein are implemented as multiple functional facilities, each functional facility may be implemented in its own manner, and not all need to be implemented in the same way. Further, these functional facilities may be performed in parallel and / or serially, if desired, and may be shared on a running computer using a message passing protocol or any other suitable method. Memory may be used to pass information between each other.

  Generally, functional facilities include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. Typically, the functionality of the functional facilities may be combined or distributed as desired in the system in which they operate. In some implementations, one or more functional facilities that perform the techniques described herein may together form a complete software package. These functional facilities may, in alternative embodiments, be configured to interact with other unrelated functional facilities and / or processes to implement the application of the software program.

  Some example functional facilities are described herein to perform one or more tasks. However, the described functional facilities and task divisions are merely examples of the types of functional facilities that may implement the example techniques described herein, and embodiments may be used to identify any of the functional facilities. It should be understood that the invention is not limited to being implemented in numbers, divisions, or types. In some implementations, all functions may be implemented in one functional facility. Also, in some implementations, some of the functional facilities described herein may be implemented with or separately from others (ie, as one unit or another unit), or these functional facilities may be implemented. It is to be understood that some of the steps may not be performed.

  Computer-executable instructions that implement the techniques described herein (when implemented as one or more functional facilities or any other method), in some embodiments, provide functionality to a medium. To be encoded on one or more computer readable media. Computer readable media include magnetic media, such as a hard disk drive, optical media such as a compact disk (CD) or digital versatile disk (DVD), Blu-ray disk, persistent or non-persistent solid state memory (eg, flash). Memory, magnetic RAM, etc.), or any other suitable storage medium. Such computer-readable media can be obtained in any suitable manner, for example, as computer-readable storage medium 3406 of FIG. 34 described below (ie, as part of computer device 3400), or as another separate storage medium. Can be implemented. As used herein, "computer-readable medium" (also referred to as "computer-readable storage medium") refers to a tangible storage medium. A tangible storage medium is non-transitory and has at least one physical structural component. In “computer-readable medium” as used herein, at least one physical structural component is a process of creating a medium with embedded information, a process of recording information thereon, or a medium of information. Has at least one physical property that can be changed in any way during any other process of encoding. For example, the magnetic state of some of the physical structure of a computer-readable medium can be changed during the recording process.

  In some (but not all) implementations of the present technology that may be embodied as computer-executable instructions, these instructions may be implemented on one or more suitable computing devices operating on any suitable computer system. Alternatively, one or more computing devices (or one or more processors of one or more computing devices) may be programmed to execute computer-executable instructions. The computing device or processor has instructions stored in a manner accessible to the computing device or processor, such as a data store (eg, an on-chip cache or instruction register, a computer-readable storage medium accessible via a bus, etc.). In some cases, it may be programmed to execute instructions. The functional equipment containing these computer-executable instructions may be one multi-purpose programmable digital computing device, two or more multi-purpose computing devices sharing processing power and performing the techniques described herein together. , A computer device or a cooperative system of computer devices (co-located or geographically distributed) dedicated to performing the techniques described herein, performing the techniques described herein. One or more FPGAs (field-programmable gate arrays), or any other suitable technology for managing these operations.

  FIG. 34 illustrates one exemplary implementation of a computing device in the form of a computing device 3400 that can do other things, but may be used in a system that implements the techniques described herein. It should be understood that FIG. 34 is not intended to represent the components required for a computing device to operate in accordance with the principles described herein, and is not intended to be a comprehensive description. is there.

  The computing device 3400 includes at least one processor 3402, a network adapter 3404, and a computer-readable storage medium 3410. Computer device 3400 is, for example, a medical device, a desktop or laptop personal computer, a personal digital assistant (PDA), a smart mobile phone, a server, or any other suitable computer device as described above. obtain. Network adapter 3404 may include any suitable hardware and / or saltware that enables wired and / or wireless communication of computing device 3400 with any other suitable computing device over any suitable computing network. Can be Computer networks include any suitable wired and / or wireless devices for exchanging data between two or more computers, including wireless access points, switches, routers, gateways, and / or other network equipment, and the Internet. Communication media. Computer-readable medium 3410 may be configured to store data processed by processor 3402 and / or instructions to be executed. Processor 3402 enables processing of data and execution of instructions. The data and instructions may be stored on a computer readable storage medium 3410.

  In embodiments where the device 3400 is a medical device as described herein, the device 3400 may include an invasive medical device 3406 that is inserted into the body of a subject to diagnose and / or treat the subject. The device 3406 includes the invasive probe 3408 described above.

  Data and instructions stored on the computer-readable storage medium 3410 may include computer-executable instructions that implement techniques that operate in accordance with the principles described herein. In the example of FIG. 34, the computer readable storage medium 3410 stores computer-executable instructions that implement various facilities and store various information as described above. Computer readable storage medium 3410 may store lesion analysis facility 3412 to analyze one or more characteristics of the lesion, including the composition of the lesion, and / or to determine a treatment recommendation based on the analysis. Computer readable storage medium 3410 may further store conditions 3414 relating to treatment options that may be used by equipment 3412. Also, the computer readable storage medium 3410 can store the learning facility 3416 and the chronicle creation facility 3418.

  Although not illustrated in FIG. 34, the computing device may further have one or more components and peripherals, including input devices and output devices. These devices can be used, inter alia, to present a user interface. Examples of output devices that can be used to provide a user interface include a printer or display screen for visually presenting the output, and speakers or other sound generating devices for audibly presenting the output. Examples of input devices that can be used for the user interface include a keyboard, and a pointing device, such as a mouse, a touchpad, and a digital tablet. As another example, a computing device may receive information entered via speech recognition or in other acoustic formats.

  Embodiments have been described in which the technique is implemented with circuitry and / or computer-executable instructions. It should be understood that some embodiments may be in the form of a method, wherein at least one example is provided. Acts performed as part of this method may be ordered in any suitable manner. Therefore. Embodiments may be constructed that perform acts in a different order than illustrated, which may indicate that several acts are performed at the same time, even though shown as a series of acts in the illustrated embodiment. May be included.

  The various aspects of the above embodiments may be used alone, in combination, or in various configurations not specifically discussed in the above embodiments, The present invention is not limited to the application of the aspect to the details and configurations of the components described in the above description or illustrated in the drawings. For example, aspects described in one embodiment may be combined in some way with aspects of other embodiments.

  The use of the terms "first," "second," "third," and the like, in a claim to modify the elements of the claim is, in itself, a single patent. It is not intended to imply any priority, advantage, or order of the claims over another claim, nor does it imply an order in time to perform an act of a method. To distinguish elements, a claim element with one particular name is used as an indicator to distinguish it from another element with the same name (but due to the use of the original term). .

  Also, the words and terms used herein are for description and should not be considered as limiting. As used herein, “comprising”, “comprising”, “having”, “including”, “including”, and variations thereof include the items listed thereafter and their equivalents, as well as additional items Means

  The term "exemplary" is used herein to mean serving as an example, instance, or illustration. Thus, by way of example, any embodiments, implementations, processes, features, etc., described herein are to be understood as illustrative examples, unless otherwise indicated preferred or preferred examples. Should not be understood.

  Thus, while several aspects of at least one embodiment have been described, it should be understood that various changes, modifications, and improvements will readily occur to those skilled in the art. Such alterations, modifications, and improvements are intended to be part of this disclosure, and are intended to be within the spirit and scope of the principles described herein. Thus, the above description and drawings are by way of example only.

Claims (60)

A method of training a system to identify at least one characteristic of a anatomy, said method comprising:
Receiving training data comprising a plurality of sets of impedance measurements for the anatomy;
Identifying, from each set of the plurality of sets of impedance measurements, a first subset of training data that includes a first subset of impedance measurements;
Identifying a first plurality of characteristics from the first subset of the identified training data, wherein the first plurality of characteristics comprises at least one at least one characteristic resulting from the first subset of the identified training data. Including derived properties, and
Training a model using at least one machine learning technique with the first plurality of identified characteristics to create a first trained model;
Including a method.   The at least one machine learning technique comprises at least one of a support vector machine (SVM) technique, an artificial neural network (ANN) technique, a k-nearest neighbor (kNN) technique, or a decision tree learning technique. The method described in.   The method of claim 1, wherein the plurality of sets of impedance measurements include electrochemical impedance spectroscopy (EIS) measurements on the anatomy.   The method of claim 1, wherein at least one characteristic of the anatomy comprises a composition of the anatomy.   The method of claim 1, further comprising using the first trained model to identify at least one characteristic of the anatomy. Using the first trained model to identify at least one characteristic of the anatomy,
Capturing at least one impedance measurement of the anatomy;
Identifying at least one characteristic of the anatomy based on the captured at least one impedance measurement of the anatomy using at least one machine learning parameter associated with the trained model;
The method of claim 5, comprising: Determining whether the first trained model matches at least one performance goal;
Using the first trained model to identify at least one characteristic of the anatomy responsive to the first trained model that matches the at least one performance goal;
The method of claim 1, further comprising: A second subset of training data including a second subset of impedance measurements from each set of the plurality of sets of impedance measurements responsive to the first trained model that does not match the at least one performance goal. Identifying and
Identifying a second plurality of characteristics from the second subset of the identified training data, wherein the second plurality of characteristics comprises at least one of a plurality of characteristics arising from the second subset of the identified training data. Including derived properties, and
Training the model using the at least one machine learning technique with the second plurality of identified characteristics to create a second trained model;
Determining whether the second trained model matches the at least one performance goal;
Using the second trained model to identify at least one characteristic of the anatomy responsive to the second trained model that matches the at least one performance goal;
The method of claim 7, further comprising:   Identifying the second subset is based at least in part on a genetic algorithm that receives as an input the first subset and a performance metric for the first trained model, 9. The method of claim 8, comprising selecting a subset.   The method of claim 1, wherein the anatomy is a lesion of an animal tract. A method of training a system to identify at least one characteristic of a anatomy, said method comprising:
Selecting a subset of impedance measurements from each of the plurality of sets of impedance measurements for the anatomy to generate a plurality of subsets of impedance measurements, each set of impedance measurements being adapted for application of a signal at a different frequency. Including an impedance measurement of one of said anatomical structures responsive;
Generating a plurality of sets of properties, wherein each set of properties characterizes a subset of the impedance measurements of the plurality of subsets, wherein each set of properties is present in at least one of the subsets of the impedance measurements. Having at least one derived characteristic arising from a subset of the characteristic and impedance measurements;
Training a model to recognize at least one feature of the target anatomy based on the input impedance measurement for the target anatomy, wherein the training comprises at least one machine learning technique; Applying to the plurality of sets of properties characterizing the plurality of subsets of impedance measurements to create a trained model;
And activating at least one processor to perform such an action. The at least one processor to process impedance measurements for a first anatomy using the trained model to determine at least one characteristic of the first anatomy. 12. The method of claim 11, further comprising actuating. Processing the impedance measurement using the trained model,
Selecting a first subset of impedance measurements for the first anatomy;
Creating a first set of properties characterizing the first subset, including the at least one property and the at least one derived property;
Processing the first set of characteristics using the trained model to determine at least one characteristic of the first anatomy;
13. The method of claim 12, comprising: capturing in vivo an impedance measurement for the first anatomy, wherein the capture is responsive to applying a different frequency electrical signal to the first anatomy. The method of any preceding claim, further comprising measuring the impedance of the structure.   The method of any preceding claim, wherein training the model to recognize the at least one feature comprises training the model to recognize a composition of a desired anatomy. Method.   16. The method of claim 15, wherein training the model to recognize the at least one feature comprises training the model to recognize one or more biological materials that make up a anatomy. .   Training the model to recognize the one or more biological materials comprises, for each subset of the plurality of subsets, training based on information about one or more biological materials of a biological structure to which the subsets correspond. 17. The method of claim 16, comprising:   The method of any one of the preceding claims, wherein training the model to recognize the at least one feature comprises training the model to diagnose the anatomy. The trained model is a first trained model,
The method comprises
Determining whether the first trained model meets performance goals;
Selecting a different subset of impedance measurements from the plurality of sets to generate a second plurality of impedance measurements in response to the determination that the first trained model does not meet the performance goal; To do
Creating a property of the plurality of sets with respect to the second plurality of subsets;
Applying the at least one machine learning technique to the second plurality of characteristics to create a second trained model;
Determining whether the second trained model meets the performance goal;
Further comprising activating the at least one processor to perform such an act.
A method according to any one of the preceding claims. The trained model is a first trained model,
The method comprises
Determining whether the first trained model meets performance goals;
Creating a second plurality of characteristics in response to a determination that the first trained model does not meet the performance objective, wherein each set of the second plurality of characteristics is Characterizes a subset of the impedance measurements of the plurality of subsets, wherein the second plurality of properties are not included in the at least one property or the at least one derived property. A characteristic, wherein said different characteristics include at least one second characteristic present in the subset of impedance measurements and at least one second derived characteristic arising from the subset of impedance measurements;
Applying the at least one machine learning technique to the second plurality of characteristics to create a second trained model;
Determining whether the second trained model meets the performance goal;
Further comprising activating the at least one processor to perform such an act.
The method according to any one of claims 11 to 18. The trained model is a first trained model,
The method comprises
Determining whether the first trained model meets performance goals;
Responsive to a determination that the first trained model does not meet the performance goal, iteratively training the model until the trained model meets the performance goal; Each iteration of training by repeating
Selecting different subsets of impedance measurements from said plurality of sets to generate different subsets of impedance measurements different from previous iterations; or generating a second plurality of sets of properties. Wherein each set of the second plurality of characteristics characterizes a subset of the impedance measurements of the plurality of subsets, wherein the second plurality of characteristics of the at least one characteristic of a previous iteration or the At least one second property present in the subset of impedance measurements and at least one second property resulting from the subset of impedance measurements, wherein the different properties include one or more properties not included in at least one derived property. Generating a second plurality of sets of properties, including the derived properties of And include one or both,
Further comprising activating the at least one processor to perform such an act.
The method according to any one of claims 11 to 18.   Claims: Iteratively training the model comprises applying a genetic learning process to the selection of impedance to include in the subset of impedance measurements and / or the selection of characteristics to include in the set of characteristics. Item 22. The method according to Item 21.   Determining whether the first trained model satisfies the performance goal may include determining whether the first trained model has an accuracy that exceeds a threshold based on an input impedance measurement. 22. The method of claim 21, comprising determining whether to accurately identify. In response to determining that the trained model meets the performance goal,
Setting up a system including a probe for measuring anatomical impedance at at least one set of frequencies used in training the trained model meeting the performance goals;
Used to determine at least one feature of the anatomy of interest, the trained model, and the set of properties used in training the trained model meeting the performance goals. Setting up a system that includes the invasive probe,
22. The method of claim 21, further comprising: From each of the plurality of sets of impedance measurements, selecting a subset of impedance measurements,
Selecting a set of frequencies;
Including each set of impedance measurements in each corresponding subset if the impedance measurements are for one of the sets of frequencies;
22. The method of any one of claims 11 or 19-21, comprising:   26. The method of claim 25, wherein selecting the set of frequencies includes evaluating, for different frequencies, the impedance measurements of the set of impedance measurements better distinguish anatomy.   26. The method of claim 25, wherein creating a plurality of sets of properties includes evaluating how different properties from a group of properties better distinguish anatomy. The at least one property present in the subset includes a value of an impedance measurement in the subset;
The at least one derived characteristic comprises a value determined from performing at least one computer operation on one or more impedance values of the subset.
A method according to any one of the preceding claims.   29. The method of claim 28, wherein performing the at least one computer processing includes performing one or more statistical analyzes on one or more impedance values of the subset.   29. The method of claim 28, wherein performing the at least one computation comprises determining a value representing a change between impedance measurements of the subset. Using at least one second machine learning technique, between an impedance measurement on the target anatomy of the animal and another anatomy of the animal that may be located near the target anatomy in the animal. Training a second model, wherein training the second model comprises using one or more impedance measurements on one or more other anatomical structures to distinguish The preceding claim, comprising training two models, further comprising: wherein the at least one second machine learning technique can be the same as, or at least partially different from, the at least one machine learning technique. The method of any one of the preceding clauses.   Training the second model using one or more impedance measurements comprises training the model using a set of properties characterizing the impedance measurements for the one or more other anatomical structures. 32. The method of claim 31, comprising, wherein the set of properties comprises the same property as the properties of the plurality of sets. The method includes selecting the one or more impedance measurements as a subset of a set of impedance measurements characterizing the one or more other anatomical structures; and selecting the set of properties from the one or more impedance measurements. Creating, and further comprising,
33. The method according to claim 32. Responsive to receiving the impedance measurement, which may correspond at least in part to the first anatomy,
Filtering the impedance measurements using the second model to remove impedance measurements that do not correspond to the first anatomy and retrieving a filtered set of impedance measurements for the first anatomy. ,
Processing the filtered set of impedance measurements using the trained model to determine at least one characteristic of the first anatomy;
The method according to any one of claims 31 to 33, further comprising: Processing the filtered set of impedance measurements using the trained model,
Selecting a first subset of impedance measurements for the first anatomy;
Creating a set of properties characterizing the first subset, the set of properties including the at least one property and the at least one derived property;
Processing the set of characteristics using the trained model to determine at least one characteristic of the first anatomy;
35. The method of claim 34, comprising:   The method of any one of the preceding claims, wherein the plurality of sets of impedance measurements include electrochemical impedance spectroscopy (EIS) measurements on the anatomy.   The method of any one of the preceding claims, wherein the anatomy is an animal lesion. A method of training a system to identify at least one characteristic of a anatomy, said method comprising:
For a plurality of sets of impedance measurements, and for each set of impedance measurements, the set of impedance measurements trains a first model using a corresponding anatomical index, wherein the plurality of sets of impedance measurements are trained. The set includes impedance measurements for a plurality of types of anatomy, wherein the training is based at least in part on the impedance measurements and trains the first model to perform an impedance measurement for a first type of anatomy. Including distinguishing from impedance measurements for one or more other types of anatomy, wherein the training includes applying at least one machine learning technique;
Training a second model by at least partially applying the at least one machine learning technique to an impedance measurement for the first type of anatomy to obtain at least one of the first type of biological material. Identifying features; and
And activating at least one processor to perform such an action. Filtering the input set of impedance measurements using the first model to retrieve a filtered input set of impedance measurements associated with the first anatomy of the first type;
Determining at least one characteristic of the first anatomy using the second model;
39. The method of claim 38, further comprising activating the at least one processor to perform the act of: Capturing an input set of the impedance measurements in vivo, wherein the capturing is responsive to the application of a different frequency electrical signal to the anatomy, the impedance of a anatomy including the first anatomy. 40. The method of claim 39, further comprising measuring. Determining a plurality of sets of impedance measurements from a plurality of input sets of impedance measurements, wherein each set of the plurality of sets of impedance measurements is a corresponding set of the plurality of input sets of impedance measurements. A subset of the anatomical structure, wherein each set of the plurality of input sets is responsive to the application of the plurality of sets of impedance measurements including signals at different frequencies and impedance measurements at only a portion of the different frequencies. Including impedance measurement,
39. The method of claim 38, further comprising operating the at least one processor as such. Training a first model using the plurality of sets of impedance measurements comprises training the first model using characteristics characterizing each set of the plurality of sets of impedance measurements. Including
Training the second model using an impedance measurement for the first type of anatomy comprises using the characteristic characterizing the impedance measurement for the first type of anatomy to generate the second model. Including training the model,
42. The method according to claim 41.   43. The method of claim 42, wherein the characteristic characterizing each set of the plurality and the characteristic characterizing the impedance measurement for the first type of anatomy are the same characteristic. Training the first model and training the second model,
At least some of the repetitions characterizing each set of the plurality and characterizing the impedance measurement for the first type of anatomy that are different from properties used in training one or more previous repetitions. 44. The method of claim 43, comprising using the characteristics to iteratively train the first model and the second model until at least one performance goal is met.   Iteratively training the first model and the second model may include using the impedance measurement for at least one frequency different from the frequency used for training in one or more previous iterations. 46. The method of claim 44, further comprising in repeating the part.   The characteristic characterizing each set of the plurality and the characteristic characterizing the impedance measurement for the first type of anatomy are different characteristics. 43. The method according to claim 42. Training the first model includes iteratively training the first model until at least one first performance goal is met, and iteratively training the first model. Using at least some of the iterations a property that characterizes each set of the plurality of sets of impedance measurements that is different from a property used in training one or more previous iterations of the training of the first model. Included in
Training the second model includes iteratively training the second model until at least one second performance goal is met, and iteratively training the second model. Using at least some of the characteristics that characterize the impedance measurement for a first type of anatomy that is different from the characteristics used in training in one or more previous iterations of the training of the second model. Including in iterations,
47. The method according to claim 46.   48. The method of any one of claims 42 to 47, wherein the property comprises a property present in an impedance measurement characterized by the property and / or a property resulting from an impedance measurement characterized by the property. A method for determining at least one characteristic of a anatomy, said method comprising:
Evaluating an impedance measurement for the anatomy using at least one trained model to determine the at least one characteristic, wherein training the at least one trained model; A method comprising activating at least one processor to perform an action such as distinguishing between anatomical structures having different characteristics. The at least one trained model is at least one first trained model;
The method further includes filtering one or more sets of impedance measurements using at least one second trained model to generate a filtered impedance measurement, wherein the received impedance measurement is Includes at least a portion of the impedance measurement corresponding to the anatomy, wherein the at least one second trained model distinguishes between an impedance measurement not corresponding to the anatomy and an impedance measurement corresponding to the anatomy. Have been trained to
Evaluating the impedance measurement using the at least one first trained model evaluates the filtered impedance measurement using the at least one first trained model. Including
50. The method according to claim 49. 51. The method of claim 50, further comprising capturing an impedance measurement of the anatomy in vivo using an insertable device having one or more sensors for detecting impedance.   52. The method of claim 51, wherein the anatomy is an animal lesion. The at least second trained model is further configured to distinguish between an erroneous impedance measurement and an erroneous impedance measurement;
Filtering one or more sets of the impedance measurements includes detecting and removing any erroneous measurements using the at least one second trained model.
The method of claim 50. Determining a method of treating a lesion in an animal, said method comprising:
Using at least one trained model to evaluate an impedance measurement for the lesion to determine a method of treating the lesion, wherein evaluating the impedance measurement comprises: Using a trained model to evaluate one or more properties of the impedance measurement, such that the one or more properties include at least one derived property resulting from the impedance measurement. Activating the at least one processor to perform the method.   Evaluating the impedance measurement using the at least one trained model includes performing one or more computations on the impedance measurement and coefficients of the at least one trained model; 55. The method of claim 54, wherein the result of the one or more computations represents a method of treating the lesion. A method of training a system to identify a method of treating a anatomy, said method comprising:
Receiving training data comprising a plurality of sets of impedance measurements for the anatomy;
Identifying, from each set of the plurality of sets of impedance measurements, a first subset of training data that includes a first subset of impedance measurements;
Identifying a first plurality of characteristics from a first subset of the identified training data, wherein the first plurality of characteristics is at least one derivative derived from the first subset of the identified training data. Including the characteristics
Training a model using at least one machine learning technique with the first plurality of identified characteristics to create a first trained model;
Including a method. A method of training a system to identify a method of treating a anatomy, said method comprising:
Selecting a subset of impedance measurements from each of the plurality of sets of impedance measurements for the anatomy to generate a plurality of subsets of impedance measurements, each set of impedance measurements being adapted for application of a signal at a different frequency. Responding, including an impedance measurement of one of the anatomical structures;
Creating a plurality of sets of properties, each set of properties characterizing a subset of the impedance measurements of the plurality of subsets, wherein each set of properties includes at least one of the plurality of subsets present in the subset of impedance measurements. Including a characteristic and at least one derived characteristic arising from a subset of the impedance measurements;
Training the model to determine a recommended treatment for the target anatomy of the plurality of treatment options from the input impedance measurement for the target anatomy, wherein the training comprises at least one machine learning technique; Applying to the plurality of sets of characteristics characterizing the plurality of subsets of the impedance measurements to create a trained model;
Actuating at least one processor to perform such an act.
Method. A method of training a system to identify a method of treating a anatomy, said method comprising:
For a plurality of sets of impedance measurements, and for each set of impedance measurements, the set of impedance measurements trains a first model using a corresponding anatomical index, wherein the plurality of sets of impedance measurements are trained. The set includes impedance measurements for a plurality of types of anatomy, wherein the training is based at least in part on the impedance measurements and trains the first model to perform an impedance measurement for a first type of anatomy. Including distinguishing from impedance measurements for one or more other types of anatomy, wherein the training includes applying at least one machine learning technique;
A second model is trained by at least partially applying at least one machine learning technique to the impedance measurement for the first type of anatomy, and the first model is selected from a plurality of treatment options. Determining a recommended treatment for the desired anatomy of the type; and
Actuating at least one processor to perform such an act.
Method. At least one processor;
59. At least one storage medium having executable coded instructions that, when executed by the at least one processor, cause the at least one processor to perform the method of any one of claims 1-58.
An apparatus, including:   59. At least one non-transitory computer-readable medium having executable coded instructions that, when executed by at least one processor, cause the at least one processor to perform the method of claim 1. Computer readable storage medium.

Images